Forecasting Construction Cost Indices: Methods, Trends, and Influential Factors

Abstract

The Construction Cost Index (CCI) is an important tool that is widely used in construction cost management to monitor cost fluctuations over time. Numerous studies have been conducted on CCI development and forecasting models, including time series, artificial intelligence, machine learning, and hybrid models. Therefore, this study seeks to reveal the complexity of CCI forecasting and identify the leading indicators, trends, and techniques for CCI prediction. A bibliometric analysis was conducted to explore the landscape in the CCI literature, focusing on co-occurrence, co-authorship, and citation analysis. These analyses revealed the frequent keywords, the most cited authors and documents, and the most productive countries. The research topics and clusters in the CCI forecasting process were presented, and directions for future research were suggested to enhance the prediction models. A case study was conducted to demonstrate the practical application of a forecasting model to validate its prediction reliability. Furthermore, this study emphasizes the need to integrate advanced technologies and sustainable practices into future CCI forecasting models. The findings are useful in enhancing the knowledge of CCI prediction techniques and serve as a base for future research in construction cost estimation.

1. Introduction

The Construction Cost Index (CCI) is an important parameter that helps in estimating construction costs at different points in time. Through the CCI, it becomes easier for the stakeholders to compare cost increases in different regions and at different times, thus helping them with planning and budgeting [1,2]. Engineering News Record (ENR) is a well-known magazine specializing in providing the latest values for the Construction Cost Index (CCI). ENR publishes its CCI monthly, which reflects the cost of specific goods and services essential to the construction industry, and it is based on the prices of construction materials like steel, cement, and lumber as well as labor wages and equipment operating and owning costs in 20 U.S. major cities [3]. Several factors influence the Construction Cost Index. Forecasting the CCI is not an easy task, as it requires the gathering of leading indicators that affect the prediction ability of the forecasting model [2]. Leading indicators give an early sign of economic changes before they happen. For example, in construction, a leading indicator would show the patterns of price changes before they happen, which can help with better planning and budgeting. That time gap between the indication and the real situation can help decide if the indicator is useful for short-term or long-term forecasting [4]. In addition, accurate cost estimation and effective project planning depend heavily on leading indicators, as they have a serious effect on construction budgets. Vital economic indicators, including inflation, interest rates, and GDP, along with others, can have a direct impact on construction cost indexes by changing labor, material, and equipment prices. Understanding these indicators will eventually lead to better CCI prediction, which can help with estimating accurate costs. These indicators can also help with choosing a proper prediction model that serves the specific needs of the analysis, whether for short-term adjustments or long-term strategic planning. This can help construction experts make informed decisions, improve cost control tools, and predict changes in market conditions that could affect project outcomes.

This research aims to address the methodologies and trends in forecasting that influence the Construction Cost Index, as well as identify the leading indicators and factors that affect the CCI. To create a structured basis for this research, the Scopus database was utilized to systematically identify and categorize relevant studies published between 2014 and 2024. Specific keywords, like “Construction Cost Index”, “construction cost forecasting”, and “construction economic indicators”, were used to incorporate the most current and important developments and advancements in the field of CCI estimation. The studies identified were carefully screened to concentrate on those most relevant to the field through a multistep filtering process to finalize the selection. The analysis started with a selection of influential and recent publications to thoroughly review key and impactful developments in the field of CCI forecasting. The selected studies were then grouped into multiple different categories, such as the type of forecasting model used, which included time-series, artificial intelligence and machine learning, and hybrid models. Furthermore, the studies were categorized based on the geographical region of the research and the leading indicators studied. After this initial review, a bibliometric analysis was utilized to illustrate the connections among essential indicators and methodologies in the research gathered from Scopus. This framework helped establish a complete understanding of the trends and challenges in CCI forecasting, forming the ground for the analyses showcased in this paper.

Research Gap and Justification

The construction industry deals with constant challenges caused by several economic, social, and market-driven factors in cost management. The greater the complexity and size of the project, the more financial risks are associated with inaccurate cost predictions. Predicting CCI trends gives stakeholders the ability to prepare for cost changes, manage risks, and allocate resources efficiently. Therefore, in such an unstable industry, detailed and accurate forecasting tools like the CCI turn out to be important for the success of project planning and execution.

A number of studies have pointed out the difficulties in forecasting the Construction Cost Index, considering the many different factors that affect those indexes. Many of these construction indexes, such as the Highway Construction Cost Index (HCCI), are affected by market volatility, leading to great challenges for accurate long-term cost estimation [5]. Traditional forecasting methods might be useful, but they often struggle with capturing the complexities of modern volatile market conditions or rapidly changing economic factors [6]. This requires an analysis of current methods and an assessment of their performance to identify areas that need further development.

This study is important for several reasons. First, it presents a thorough systematic assessment of the leading economic indicators that influence construction financials, giving a comprehensive understanding of the important variables needed for CCI forecasting. Second, it points out the gaps in current forecasting techniques, stressing the need for models that can better represent the volatility and dynamic characteristics of the construction field. Finally, the study contributes to the advancement of more resilient and flexible forecasting techniques that can be applied across different construction environments.

2. Methodology

This paper aims to discuss the CCI forecasting methods that influence how the indexes are predicted in different scenarios. Furthermore, it aims to identify the leading indicators that affect the process of CCI forecasting and explore the trends of the CCI, which include how the CCI has changed over time as a response to different conditions in the construction industry. The proposed methodology consists of a two-step approach:

• Research collection and review: This step consists of conducting an intensive publication search using the Scopus database to gather all the relevant studies on the CCI based on selected keywords to target the most up-to-date and critical studies in the field. Then, a data cleaning process is initiated to ensure the quality and relevance of the gathered records. All the steps involved in this process are documented using a PRISMA flowchart, which provides a clear visualization of the data collection process. Following this search, several publications are selected to review the advancements in CCI forecasting modeling. These studies are manually selected, as some of them are impactful in the field of CCI forecasting and others are recent publications. These studies are categorized based on the type of forecasting model used, such as time-series, AI, machine learning, hybrid models, and other advanced models. In addition to model categorization, the regional context of the studies and the vital economic and market-leading indicators in forecasting are also taken into account.
• This structured review allows for the identification of the main trends, difficulties, and developments in the field of CCI forecasting. It also provides a good starting point for determining which forecasting methods are best suited for different situations, laying the groundwork for the following stage of the analysis.
• Bibliometric Analysis: Building on the findings from the structured review and using the data retrieved from Scopus over a range of 10 years, a bibliometric review is conducted to map the relationships between different CCI indicators and modeling techniques. This analysis offers a comprehensive view of the interconnections between the research areas, the advanced forecasting techniques, and the key contributors in the field. Furthermore, it assists in addressing the patterns that provide valuable insights into how some forecasting techniques prevail in certain conditions and under the effect of specific indicators, supporting the decision-making process for all parties in the construction industry.
• The integration of the research review and bibliometric analysis allow for a solid evaluation of each model’s effectiveness and efficiency by analyzing its strengths and weaknesses in several construction scenarios, including changes in material costs, regional economic conditions, and supply chain disruptions. By assessing the connections between forecasting techniques and the leading indicators found in the bibliometric analysis, the optimal methods for particular forecasting requirements in the construction industry can be established. This method guarantees that the evaluation underscores the leading approaches and provides knowledge about their practical feasibility in actual construction cost management.

3. Data Collection

Identifying the resources needed for the research was straightforward and comprehensive, as the goal was to gather all the references that were crucial to form a good base for the study take-off.

The process started with searching Scopus, a widely recognized and comprehensive research repository known for its extensive coverage of peer-reviewed literature across various disciplines. Scopus is particularly valuable for its robust indexing of academic articles, conference proceedings, and patents, which makes it a perfect source for obtaining quality references in the field of cost index forecasting, especially in the construction industry [7]. The search was performed using specific keywords that would help narrow down the tremendous amount of information and data to capture only the core concepts that were needed for the study.

The search timespan was set to include all the publications between 2014 and 2024 to ensure the capturing of the most recent and relevant research in the field. This period was chosen to reflect the latest advancements, trends, and developments in construction cost estimation and related construction indexes. The deployed set of keywords included but was not limited to “construction cost estimation”, “Construction Price Index”, “construction inflation”, “Construction Cost Index”, “construction supply chain”, “construction economic indicators”, “construction forecasting”, “construction supply chain disruptions”, “construction cost trends”, and “construction cost estimation indexes”.

A PRISMA flow diagram was used to make things clearer and more reliable. This diagram visually shows the process of finding and selecting the references that were used in this research, making it transparent and much easier to digest. It also helped with spotting any biases or gaps in the selection of resources, ensuring that the study is credible and trustworthy [8] (see Figure 1). The total number of references gathered was 143,543. To account for a wider crowd, 8096 records were excluded, as their main language was not English. The 135,447 records remaining entered a three-part high-intensity screening process, which began with limiting the publications to three subjects, which were engineering, management and accounting, and decision sciences, which yielded a total of 41,071 publications that entered the next part of the screening process. This focused approach was crucial to make sure that the remaining records were related to the main areas of the research.

The second screening process involved the removal of records that represented source titles that were not related to the focus of this research. These source titles included “Energies”, “Tunnelling and Underground Space Technology”, “Energy and Buildings”, “Energy”, “IEEE Access”, “Geotechnical Special Publication”, and “Sustainability”.

A total of 28,813 records were left for the third and final screening process, which focused on excluding the keywords that did not accurately represent the research objectives. The process involved a highly focused review of the keywords listed on Scopus to remove the ones that were too broad or unrelated, such as “carbon dioxide”, “greenhouse gases”, “stiffness”, “railroads”, “offshore oil well production”, “piles”, “fly ash”, “lime”, “soil testing”, and “emission control”. By the end of the third screening process, the study was left with 19,284 records that met the set criteria for relevance and quality. These records formed the foundation for initiating the bibliometric review process, which studied the distribution of citations, co-authorship networks, and field trends within the proposed research.

Figure 2 shows the number of published documents in this field from the end of 2014 to 2024. It can be noticed that research efforts gradually increased every year, which indicates that the understanding of construction cost estimation is developing through the introduction of new technologies and the circulation of new ideas regarding elevating the quality and success of the CCI calculation and estimation process.

The different types of publications used in this study and their different classifications are represented in Figure 3. It shows the efforts carried out by the researchers to reach a point of understanding of the good implementation of the Construction Cost Index and construction cost management.

With the data collected as outlined above, the following section provides a review of forecasting techniques relevant to CCI prediction, forming the basis for the subsequent analysis.

4. Overview of CCI Forecasting Efforts

The CCI calculates the changes in costs of construction work and materials within a certain period. Economic conditions, labor productivity, technological changes, and market conditions are among the factors that affect CCI forecasting [9]. However, each study that has been conducted in this field has leading indicators and affecting factors that helped with developing better and more reliable models that advanced the prediction of construction cost indexes.

Forecasting the CCI is crucial to ensure proper control of the project execution. Support vector machine (SVM), neural networks (NNs), and smoothing techniques were incorporated for predicting the CCI [10]. The authors found that the CCI depends on several explanatory factors, including the Consumer Price Index (CPI), employment in construction, housing starts, money supply, the Project Price Index (PPI), GDP, building permits, and crude oil prices. They compared the techniques using the errors obtained in each model. These errors include mean square error (MSE), root mean square error (RMSE), and mean absolute percentage error (MAPE). Based on the error comparison, the prediction can be analyzed, and the best one will be applied for further CCI forecasting. Using correlation methods to determine the variables of importance, they discovered that the use of smoothing techniques yielded better results than other forecasting methods, thus safeguarding quantity surveyors from under- and overestimation. This approach highlights the importance of having proper and up-to-date prediction models to help control the cost of a project and the flow of the budget. In a recent study, the SVM model was utilized to forecast the CPI, as it is a very important indicator for the forecasting of the CCI. The model was built with the help of the short vector regression (SVR) to be able to apply linear regression in multidimensional feature space [11]. It was found that the SVM model yielded better results than the other AI model, which was a radial basis function neural network (RBFNN). The comparison was based on the error metrics, where the SVM scored lower errors, deeming its reliability for indicator forecasting as an AI model. The study noted the limited availability of long-term CPI data, which may affect the performance of the forecasting models that depend on such data.

Another proposed forecasting technique is based on using seasonal auto-regressive integrated moving average (SARIMA) models [12]. In their research, the authors found that SARIMA models showed high accuracy in predicting the CCI, beating the predictions of ENR subject matter experts. However, they pointed out that the SARIMA model had difficulties with the discrete jumps in the CCI and recommended that future studies should consider using the Poisson procedure to model and account for such jumps.

Moreover, a study was conducted to find the models that can forecast city-level CCIs, where their combined average forms the value of the national CCI. The study compared four linear forecasting models, including different sets of ARIMA and VECM models [13]. The authors used different leading indicators in their study, including city CPI and national CPI, as well as crude oil price and M2 money supply, which reflects the broader economic liquidity. Also, it included the unemployment rate, housing starts, and federal funds rate. The results indicated that the VECM model outperformed the other models, especially when it came to dealing with city-specific indicators like the city CPI. However, the lack of available city data for the leading indicators limits the application of complex models like the VECM. This is where the ARIMA model with the Bayesian Information Criterion (BIC) becomes effective for forecasting CCI values.

Furthermore, a modified K-nearest neighbors (KNN) algorithm was used to predict the ENR’s CCI [14]. The process started by collecting the data for several identified economic variables. Then, the data were split into training and testing sets, and prediction models were developed based on these sets. The authors used the Granger causality test to determine the leading indicators for the CCI. The Consumer Price Index (CPI), crude oil prices, GDP, and building permits were the leading indicators for their research. Even though the KNN algorithm seemed to work well, it required adjustment to address the jumps in CCI and improve the prediction accuracy. The study recommended that more leading indicators should be added and lag periods should be changed to obtain better forecasting results. The study also indicated that the model did not capture a jump in data and recommends the development of more advanced machine learning algorithms to capture jumps in data, which can help with enhancing CCI prediction.

Machine learning was employed to predict the CCI for the short, medium, and long terms [15]. The two machine learning algorithms used were K-nearest neighbors (KNN) and perfect random tree ensembles (PERT), which are utilized to enhance CCI forecasting, especially in the mid- and long-term. It was found that their methods yielded a higher prediction accuracy than the conventional time-series models. The study indicated that there is a need to enhance long-term CCI prediction for better project budget and cost management. They noted that the major drawback of their methods was that the process of building the prediction model was rather time-consuming, but once the model was built, it could be applied with only minor adjustments. This study shows that machine learning algorithms can be applied to predict CCI with high accuracy and reliability for different time horizons. Moreover, a recent study was conducted to investigate several machine learning models in CCI prediction [16]. These models included linear regression, SVM, and neural network models. The authors used historical cost and market environment data as inputs for the forecasting models. The study was based on three cases where neural networks showed the highest accuracy among all the models. However, the authors indicated that there is still room for improvement, especially with better data quality and model refinements, as well as real-time integration, which limits the model’s responsiveness to sudden market changes.

A visibility graph network was also implemented for forecasting the CCI; it transforms time-series CCI data into nodes on a graph, where each node represents a CCI data point. Once the network is established, the node degree, which represents the number of connections a node has, is calculated [17]. This degree has an important role in assigning weights to each node to help predict future CCI values. The study compared this method with other well-known forecasting methods, including simple moving average (SMA), ARIMA, SARIMA, vector autoregression, KNN, and PERT. The visibility graph was tested for short-, mid-, and long-term CCI predictions. The authors found that the visibility graph outperformed all the other models in every term length, demonstrating its superior accuracy. However, the authors acknowledge that while their method excels in prediction accuracy, it currently does not incorporate external indicators that have been shown to influence the CCI, which points to the need for more complete models that can account for all the indicators that can affect the forecasting process.

The variability of the Highway Construction Cost Index (HCCI) was addressed with the help of long short-term memory (LSTM) neural networks [5]. They identified HCCI uses as a cost inflation factor, market indicator, purchasing power indicator of federal or state agencies, and a comparison tool between different regions. Texas HCCI data were used in the modeling process, and the results were compared to those from the SARIMA model. In their work, the authors established that LSTM models were useful in estimating HCCI volatility in all three forecasting scenarios: short-term, medium-term, and long-term prediction, which is useful in cost and risk management. This research proved that the use of enhanced neural networks is suitable for identifying the interactive effects and making precise predictions. The results indicate that the use of more complex neural networks is crucial for controlling the fluctuation in construction cost indexes.

The LSTM neural network model was also used to forecast China’s CCI [18]. The authors listed 16 leading indicators, such as economic, financial, energy, and market indicators. The developed LSTM model was found to perform great because of its ability to process high-dimensional vectors and record historical information. Also, compared to other advanced models, like support vector machine (SVM), the LSTM model performed better in short-term forecasting. Their study also supported the use of LSTM neural networks in the construction of cost indexes and recommended the inclusion of more variables to capture market changes. This study also revealed that neural networks are effective in generating precise and credible CCI forecasts for the construction sector.

In Egypt, artificial neural network (ANN), linear regression, and autoregressive time-series models were compared to see which is better for predicting the CCI [6]. The study focused on concrete structures, with the use of key materials, including steel, cement, bricks, sand, and gravel. The performance of the forecasting models was then evaluated using the mean squared and the mean absolute errors. The research stated that the autoregressive time-series model was the most appropriate since it had the lowest error average of the models, while ANN came in second place. Linear regression was not realistic, as it yielded linear predictions. This study also showed that time-series analysis is effective in producing accurate CCI predictions and that it is crucial to update the prediction models to increase their effectiveness.

Hybrid models were another research field that kept on improving day by day, as they combined the best of multiple techniques to overcome some of the problems that other models could not. One of these hybrid models is the unique self-adaptive structural radial basis neural network intelligence machine (SSRIM), which helped engineers deal with the high variability of the CCI and its indicators [9]. In this model, multivariate adaptive regression splines (MARS) are used to examine several leading indicators (factors) based on their relative importance. The highly important factors are then used as input variables in the radial basis function neural network (RBFNN) to forecast CCI values. During this process, the artificial bee colony (ABC) algorithm is used to select the best variables, which are now input parameters, of the RBFNN to maximize the predictive accuracy of the model. As promised, the model yielded low average error numbers, which is a great indicator of the success of this technique.

The literature shows that the CCI is an important aspect of any construction project, but its forecasting process differs depending on the project and the region. Certain studies pointed to the potential inaccuracies of CCI-based estimates and called for methodological standardization techniques and future investigations. The application of AI and machine learning has proven useful in improving CCI prediction, but there is a difficulty in identifying the discrete jumps and long-term trends. The literature shows that although basic models like SARIMA can be used as a safe starting point, the use of more sophisticated methods, such as LSTM and hybrid models, can greatly improve the accuracy of the predictions.

5. Bibliometric Analysis

Bibliometric analysis is a technique that enables researchers to summarize large amounts of research and reference data into clusters where the data are visible and organized in a related manner. It provides an opportunity to analyze the aspects of a specific field and gives an idea of the emerging research areas in that field [19]. Furthermore, it forms a complete and comprehensive picture of the current situation of a specific industry. It reveals the trends, patterns, and impacts within a particular body of literature or research area. Also, through different analysis procedures, this approach can identify the most influential authors, documents, and topics; track the research development in time; and uncover research cooperation efforts.

The bibliometric analysis method is based on discussing the ideas of previous and current papers in an interaction map where information about relationships between keywords and research can be viewed. This analysis can be carried out through a unique user interface like VOSviewer (v 1.6.18), a software created for building and visualizing bibliometric networks. Its flexibility and user-friendly interface make it a potent tool for bibliometric analysis, providing researchers with several advantages for studying the structure and development of scientific fields [20]. The software uses advanced clustering algorithms to group similar items like articles, journals, and keywords. This feature helps determine the major research areas and their relative scopes. Density mapping emphasizes the main research areas, illustrating the spots where scholarly activity is focused [20].

Therefore, the intuitive visualization tool of VOSviewer improves the interpretability of complex bibliometric data, enabling more profound insight into the relationships and the intellectual structure of a research field. As such, using VOSviewer in bibliometric research provides a complete structure for analyzing and visualizing the development and influence of scientific research, which helps authors and researchers make sense of a large amount of academic literature more efficiently.

5.1. Overview of Quantitative Analysis

The data gathered in the previous steps were implemented in VOSviewer and analyzed in three main stages. Keyword co-occurrence was the first stage of the bibliometric analysis, which outlines the major research interrelationships and identifies research challenges and areas of potential future research based on the keywords found in each document inside the data. Analysis of the frequency of co-occurrences of keywords from literature allows researchers to visualize the intellectual structure of a field effectively. The map generated from this analysis shows the main trends and themes. It demonstrates the relationships between them, which gives a graphic representation of how different areas of studies interconnect and intersect.

Then, a co-authorship analysis was conducted to show the research contributions in the academic community, determining the research work’s geographical distribution. This analysis can reveal the pattern of international collaboration efforts, which identifies the influential authors and countries in the global research network. Such analysis can reveal the main countries that contribute to the study field, offering perspectives on the scientific standing of these countries. The bibliometric analysis was completed with a citation analysis, which evaluated the importance and impact of the most influential publications and sources. This analysis allows for the tracking of how often and where a work is cited, identifying the most influential papers, authors, and journals, which emphasizes the foundational and vital contributions in a field.

5.2. Co-Occurrence Analysis

Before generating the analysis map, cleaning the data was required, as many keywords were similar, affecting the map’s quality and validity [20]. For instance, keywords like “supply chain”, “supply chain management”, and “supply chain network” were merged under “supply chains”, which displayed a better connection with all other keywords. Also, “costs”, “cost estimations”, and “cost estimation” were gathered under the umbrella of “cost estimating.” The minimum number of occurrences of a keyword was set to 25 occurrences and above, leaving 512 keywords that met that criterion. However, only 28 keywords were used to base the study on, as they were the only ones that met the research criteria.

Upon generating the map, it was noticed that it consisted of five main clusters consisting of 28 keywords, 212 links, and a total link strength of 1688. Each cluster had a color to serve as a visual cue, simplifying the identification of different areas of the map and representing a group of similar data and keywords, which are critical for understanding the thematic structure of the studied publications and their keywords. All the keywords inside these clusters were connected using 212 links; this indicates the vast points of contact that these keywords have in common (see Figure 4).

The network shows a complete representation of how various cost indicators and forecasting methods are connected, which highlights the relationships between key subjects like supply chains, monetary policy, machine learning, forecasting, and construction costs. The network also shows the keywords as nodes, some of which are bigger than others. The size of the node indicates the number of times a keyword occurred in the literature used in the bibliometric analysis.

Another way to view keyword relationships is to identify the most mentioned tools and methods by examining a density visualization map. This map forms a special aura around the factors, indicating the cluster’s strength of connections or co-citations. A higher density might suggest a well-established research area with many successful studies. In contrast, a lower density could indicate an emerging field or a less explored topic (see Figure 5). Compiling network clusters can help researchers understand how a study field emerges and evolves [19].

The color intensity of each point in the density visualization map indicates the keyword’s occurrence in that area. Colors range from blue to green to yellow. The larger the number of items in a point’s vicinity and the higher the weights of the nearby items, the more yellow the hue of the point is [20]. The map also depicts the importance of cost estimation in reference to the other keywords and factors shown where the yellow color is intensely concentrated, revealing the significance of cost estimation for connecting these keywords.

Table 1. Number of occurrences and features of keywords.

№	Keyword	Occurrences	Links	Total Link Strength	Avg. Pub. Year	Avg. Citations
Cluster 1
1	Supply chains	452	18	258	2020.93	19.5
2	Supply-chain disruptions	210	9	211	2020.77	31.54
3	Blockchain	81	11	38	2021.77	18.25
Cluster 2
1	Monetary policy	528	14	137	2020.03	3.68
2	GDP	359	18	121	2020.08	5.13
3	Inflation	245	23	196	2020.61	3.57
4	Exchange rate	153	16	108	2020.14	4.2
5	Trade openness	97	4	12	2020.25	7.6
6	Money supply	77	12	70	2020.45	2.73
7	Interest rates	67	15	52	2020.21	3.46
8	VECM	38	8	17	2019.63	3.84
Cluster 3
1	Forecasting	577	26	554	2020.4	9.49
2	Crude oil price	158	15	86	2019.22	11.56
3	Time series	113	20	160	2020.25	11.54
4	Consumer Price Index	68	14	66	2021.06	2.91
5	ARIMA	61	17	65	2020.87	3.33
6	Long short-term Memory	60	19	92	2021.95	6.23
7	Price index	50	12	35	2020.46	4.08
8	Support vector machine	31	10	31	2020.68	18.55
Cluster 4
1	Construction costs	809	16	204	2019.64	6.67
2	Cost estimating	137	16	156	2020.23	10.31
3	Budget control	121	18	125	2019.83	8.12
4	Neural networks	118	18	152	2020.11	11.63
5	Cost indexes	36	7	16	2019.78	3.36
6	Operating costs	25	11	20	2019.44	11.64
Cluster 5
1	Unemployment	669	13	60	2019.78	9.63
2	Machine learning	264	25	233	2021.85	8.36
3	Artificial intelligence	148	19	101	2020.91	26.01

shows the number of occurrences, the number of links, the total link strength, the average number of citations, and the average publishing year for each of the keywords generated in the bibliometric map. The number of links is the number of connections that a topic or a publication has with others in the network, while the total link strength represents how connected and close a node or a keyword is to the subjects of the bibliometric analysis. The average number of citations is the average number of times papers related to a topic have been cited by others, while the average publishing year is the average year when the papers or studies related to a topic were published.

It can be noticed that the most common keywords include “construction costs”, “forecasting”, “supply chains”, “unemployment”, and “monetary policy”, which indicates a strong research focus on the subject of Construction Cost Index prediction. The table also includes many of the economic indicators of construction, such as “GDP” and “inflation”, which reflect the efforts to use macroeconomics to improve forecasting efficiency in projects.

Each of the clusters indicates similar keywords that together form an understanding of a specific area of research. Accordingly, the clusters in this study represent five main fields, namely, (1) supply chain management, (2) economic indicators, (3) forecasting and data science, (4) construction costs and project management, and (5) technology and AI integration.

5.2.1. Supply Chain Management (Yellow Cluster)

Supply chain plays a vital role in forming the Construction Cost Index (CCI) due to its impact on the availability and prices of construction materials. Global supply chain disruptions, whether from geopolitical tensions, natural disasters, or pandemics, can cause huge changes in material costs [21,22], which will affect the CCI.

COVID-19 is a big example of the vulnerability of global supply chains, as widespread disruptions led to material shortages and increased prices [22,23]. These disruptions inevitably had a hard effect on industries dependent on consistent supply chains, likely contributing to fluctuations in the Construction Cost Index (CCI). Adjustments in cost estimates were needed for such disruptions in ongoing and planned construction projects, showing the importance of sturdy supply chain management in maintaining a stable CCI.

Blockchain technology has surfaced as a promising solution to improve the transparency, traceability, and efficiency of digital transactions within supply chain management [24]. It also provides a reliable infrastructure for information management during all life-cycle stages of construction projects [25]. This is possible through the introduction of a decentralized platform for tracking materials, goods, and transactions to help stabilize costs by reducing fraud risk and ensuring the legitimacy of supply chain data [24,25,26]. This stability in supply chain management can indeed help with enhancing the predictability of the CCI, as the fluctuations in the policies and prices will be limited, allowing construction experts to produce more reliable cost estimates. The adoption of blockchain technology across various industries, including construction, can mitigate risks associated with supply chain disruptions. By improving transparency and traceability, blockchain can help ensure that projects remain on budget and on schedule, addressing key challenges in construction management.

Table 2. Keywords of the supply chain management cluster.

№	Keyword	Occurrences	Links	Total Link Strength	Avg. Pub. Year	Avg. Citations
1	Supply chains	452	18	258	2020.93	19.5
2	Supply chain disruptions	210	9	211	2020.77	31.54
3	Blockchain	81	11	38	2021.77	18.25

below indicate the keywords in the supply chain management and their important parameters.

5.2.2. Economic Indicators (Green Cluster)

GDP, inflation, monetary policy, unemployment, interest rates, trade openness, money supply, and exchange rates are considered leading macroeconomic indicators and critical factors of the CCI [4]. They influence the cost of materials, labor, equipment, and other resources, thereby affecting the accuracy of Construction Cost Index estimates.

As construction activity increases, the GDP growth rate also increases, which drives up the demand for materials and labor [27]. This economic expansion can drive up the demand for materials and labor, which in turn may lead to a higher CCI. Typically, the construction industry experiences inflationary pressure with increased demand for goods and materials, pushing CCIs up and making it harder to complete the cost estimation process. Interest rates are set by central banks to control inflation, which in turn affects construction expenses [28]. When monetary policy is too lenient, inflation increases and higher interest rates are required to reduce it [29]. Two key types of inflation affect the CCI: (1) demand–pull and (2) cost–push. Demand–pull inflation occurs when the demand for construction services exceeds the supply, leading to an increase in the price of construction materials and the workforce, as observed in the post-pandemic period [30]. Cost–push inflation is caused by an increase in production costs, including wages and materials, which are then transferred to the consumer [31].

Monetary policy, which may include raising interest rates, is implemented by some central banks to control inflation. By raising interest rates, central banks can reduce demand within the economy [32], which may contribute to the stabilization of the CCI by curbing the inflationary pressure on labor, materials, and equipment costs.

Exchange rates play a vital role in the construction industry, especially in countries that rely on imported materials. The weaker the domestic currency is, the higher the price of imports, which means future inflation and output growth [33]. This will lead to an increase in the cost of materials, which impacts the CCI. The degree of a country’s relation with the global economy, or trade openness, is another important indicator that can influence features of inflation dynamics by either enhancing competition through imports or exports, or by exposing the country’s economy to global price volatility [34]. This leads to fluctuations in the domestic construction industry, affecting the local CCI.

The unemployment rate has a direct and strong impact on inflation and labor costs in the construction industry. As the rate increases, it tends to lower labor costs as the supply of available workers exceeds demand, leading to lower inflation rates [35], which can result in lower CCIs. Conversely, lower unemployment rates can drive up labor costs as demand for skilled workers increases, increasing the inflation rate [36]. This escalation in labor costs can contribute to higher construction cost indexes (CCIs).

These up-and-down dynamics are relevant in regions experiencing economic recovery, where construction work tends to concentrate, leading to higher labor demand and increased wages. As wages rise, especially in areas of labor shortage, it results in higher project costs and difficulty in estimating construction costs [37].

Money supply is a tool that represents the amount of money in a nation’s economy. In the United States, there are at least two types of money supply that are measured, which are M1 and M2 [1]. M1 includes cash, checkable deposits, and traveler’s checks, while M2 includes M1 plus savings deposits, certificates of deposits, money market funds, and small denomination time deposits [1].

Rapid growth in these numbers can cause inflation, which will affect the value of the CCI. Therefore, accurate macroeconomic management is essential for economic forecasts [4]. These forecasts are crucial for predicting changes in various economic indicators, which can significantly impact the CCI and project budgets.

Table 3. Keywords of the economic indicators cluster.

№	Keyword	Occurrences	Links	Total Link Strength	Avg. Pub. Year	Avg. Citations
1	Monetary policy	528	14	137	2020.03	3.68
2	GDP	359	18	121	2020.08	5.13
3	Inflation	245	23	196	2020.61	3.57
4	Exchange rate	153	16	108	2020.14	4.2
5	Trade openness	97	4	12	2020.25	7.6
6	Money supply	77	12	70	2020.45	2.73
7	Interest rates	67	15	52	2020.21	3.46
8	VECM	38	8	17	2019.63	3.84

outlines the keywords found within the economic indicators cluster and highlights their important attributes.

The cluster also included the vector error correction model (VECM), which is an advanced model that is used to study and analyze the long-term relationships between the previous economic indicators and CCIs. The VECM proved to be more accurate than the cointegrated vector autoregression model, as a study was conducted using the United States’ Consumer Price Index (CPI), which is an important indicator in the CCI forecasting process [38]. The VECM offers valuable insights into how these factors impact cost trends over time.

5.2.3. Forecasting and Data Science (Red Cluster)

Forecasting is a crucial aspect of construction cost management, as it allows for predicting and managing changes in the construction cost index. Accurate forecasting is important for effective estimation, budgeting, risk management, and resource allocation in construction projects. Forecasting models such as ARIMA have been widely used to predict trends [39]. This model has been widely utilized to forecast Construction Cost Index (CCI) trends, providing valuable insights into future cost fluctuations. ARIMA tends to work better in short- and mid-term forecasting, especially when the data do not show a seasonal pattern [40].

Machine learning has significantly improved the forecasting accuracy. Advanced machine learning models, such as long short-term memory (LSTM) models, are effective at capturing complex patterns in data that traditional models might overlook [18]. The LSTM model helps address the issues of long-term dependencies in neural network data, as it is capable of retaining a lot of information over long periods of time [41]. This makes the LSTM very useful for CCI prediction

Support vector machine (SVM) is another machine learning technique that has been used for forecasting and predicting CCIs. SVM is great at handling non-linear relationships between variables using linear modeling, which makes it useful in complicated forecasting situations where budgeting is vital in construction projects [42].

The cluster also includes some important leading indicators that crucially affect the forecasting process, such as crude oil price, the Consumer Price Index (CPI), and the price index, which can directly impact economic measures like inflation. The increase in crude oil prices can have a positive impact on inflation [43]. Similarly, the Consumer Price Index (CPI) is frequently used as an indicator for forecasting the CCI, as it controls parts of the economy, which can be crucial for the modeling process [9,10,14]. The keywords in the forecasting and data science cluster, along with their key characteristics, are shown in

Table 4. Keywords of the forecasting and data science cluster.

№	Keyword	Occurrences	Links	Total Link Strength	Avg. Pub. Year	Avg. Citations
1	Forecasting	577	26	554	2020.4	9.49
2	Crude oil price	158	15	86	2019.22	11.56
3	Time series	113	20	160	2020.25	11.54
4	Consumer Price Index	68	14	66	2021.06	2.91
5	ARIMA	61	17	65	2020.87	3.33
6	Long short-term memory	60	19	92	2021.95	6.23
7	Price index	50	12	35	2020.46	4.08
8	Support vector machine	31	10	31	2020.68	18.55

below.

5.2.4. Construction Costs and Project Management (Blue Cluster)

Accurate construction cost estimation depends on understanding and predicting fluctuations in CCIs, which are affected by factors like material, labor, and equipment costs, along with overhead expenses. This cluster gathered the construction financial aspects together, as it included “construction costs”, which was the most occurring keyword, with 809 occurrences, showing its importance in the cohesion of the other keywords and clusters in the field of CCI forecasting. Also, terms like “cost estimating” and “cost indexes” define the CCI prediction process, as they include many of the indicators that are considered crucial foundations for effective forecasting modeling.

Another vital aspect that directly affects the CCI is operating costs. Fluctuations in maintenance and energy costs, for instance, can lead to significant changes in operating costs [44]. Budget control is a great example of CCI forecasting advantages. Effective CCI prediction leads to better budget control in construction projects, starting from preparing more reasonable financial offers to estimating project budgets that may take a long time to prepare using traditional ways [6]. Budget control requires a thorough understanding of available and future cost data, which can be greatly improved by using the latest CCI data.

Neural networks represent the mesh that can bring together the CCI factors, including the leading indicators in this cluster, to model and forecast the CCI based on their historical data. These networks include many known modeling techniques, which have been mentioned throughout the research, such as LSTM, ANN, and hybrid systems, which have multiple parameters that can affect the CCI prediction process, depending on the region, the availability of data, and the indicators used in the modeling process [6,18].

Table 5. Keywords of the construction costs and project management cluster.

№	Keyword	Occurrences	Links	Total link Strength	Avg. Pub. Year	Avg. Citations
1	Construction costs	809	16	204	2019.64	6.67
2	Cost estimation	137	16	156	2020.23	10.31
3	Budget control	121	18	125	2019.83	8.12
4	Neural networks	118	18	152	2020.11	11.63
5	Cost indexes	36	7	16	2019.78	3.36
6	Operating costs	25	11	20	2019.44	11.64

below presents the keywords associated with the construction costs and project management cluster along with their occurrences, links, and other vital features.

5.2.5. Technology and AI Integration (Purple Cluster)

This cluster represents the facade of innovation in construction cost management. The incorporation of artificial intelligence (AI), with 148 occurrences and 19 links, and machine learning, with 264 occurrences and 25 links, can point towards the connection that these tools have with the transformation that is happening in the construction industry from traditional practices to more efficient and accurate models and processes.

AI and machine learning are being widely used to improve CCI forecasting models. The perk of these models is that they can analyze huge amounts of data, identify patterns, and make predictions that were unachievable with the traditional methods [9,45]. Machine learning, powered by AI, can forecast trends based on a combination of different aspects, including historical data and different indicators, as it enables computers to learn from experience and repetitive tasks [46,47]. When applied to construction cost indexes (CCIs), these advanced techniques can provide experts with more detailed and accurate cost estimates. They can help minimize the uncertainty in the forecasting process in different scenarios, making it effective for project management and budget control [15].

The cluster also includes “unemployment”, with 669 occurrences, indicating the importance of this leading indicator in the forecasting process of the CCI. It mainly affects inflation, meaning that if unemployment decreases, inflation will increase, and vice versa. This all starts with the increasing demand for workers, especially skilled labor, which increases the demand for a wage increase, which spikes inflation, as the money supply is increased to cover these needs [36]. However, if the demand for workers is falling, it means an increase in unemployment, which means less wage demand and a lower inflation rate [48]. These inflation fluctuations will affect the modeling process for CCI forecasting, as it affects many other indicators.

Table 6. Keywords of the technology and AI cluster.

№	Keyword	Occurrences	Links	Total Link Strength	Avg. Pub. Year	Avg. Citations
1	Unemployment	669	13	60	2019.78	9.63
2	Machine learning	264	25	233	2021.85	8.36
3	Artificial intelligence	148	19	101	2020.91	26.01

below presents the keywords related to the technology and AI integration cluster along with their significant elements.

5.3. Co-Authorship Analysis

5.3.1. Author Productivity

The authors with the highest number of cited documents completed individually or through collaboration were identified to evaluate the active authors in this field. The minimum number of documents per author was set to 5 documents and the minimum number of citations was set to 20 citations. The number of documents, citations, and average publication year for each author are shown in

Table 7. Most cited authors.

№	Author	Documents	Citations	Avg. Publication Year
1	Sawik T.	20	539	2018.85
2	Ozili P.K.	8	134	2022.5
3	Moghayedi A.; Windapo A.	7	12	2020.43
4	Ghosh S.	6	14	2021.5
5	Omodero C.O.	6	41	2020.83
6	Hagemann H.	5	13	2019

.

Dr. Tadeusz Sawik had the most published documents, with 20 documents and 539 citations across all papers, with an average publication year of 2018.85, indicating sustained and consistent contributions in recent years. Dr. Sawik is an active researcher at the University of Kraków, with many research interests, like supply chain, logistics, and distribution management. Peterson Ozili came in second place, with 8 documents and 134 citations. His research interests include digital finance and development. Some of his most influential works include studies on the impact of COVID-19 on the global economy and the role of inflation on the prices of materials and goods, which came in handy in this study.

Dr. Alireza Moghayedi’s and Dr. Abimbola Windapo’s combined efforts resulted in 7 documents, with a modest number of 12 citations across the documents. Both of them are affiliated with the University of Cape Town, where Dr. Moghayedi specializes in construction management and transportation engineering, while Dr. Windapo focuses on construction business and project management. Their collaboration, with an average publication year of 2020.43, indicates emerging and ongoing research efforts in these areas. Moreover, Dr. Santadas Ghosh and Dr. Cordelia Omodero had 6 documents each; however, Dr. Omodero’s work achieved greater recognition, with 41 citations compared to Dr. Ghosh’s 14 citations.

Lastly, Dr. Harald Hagemann contributed 5 documents with 13 citations, with an average publication year of 2019, reflecting steady contributions to the field. His primary focus areas include macroeconomics and innovation policy, which helped with shaping the main aspects of the factors influencing construction cost indexes (CCIs)

5.3.2. Country Contributions

The second phase of the co-authorship analysis reviewed the number of documents and citations based on countries worldwide. The minimum number of documents per country was set to 35 documents and the minimum number of citations was set to 30 citations. The top 10 productive countries are provided in

Table 8. Country contribution.

№	Country	Documents	Citations	Total Link Strength
1	United States	3086	20888	637
2	China	2121	7414	410
3	India	1953	1003	196
4	United Kingdom	1346	232	528
5	Germany	810	808	253
6	Australia	636	197	260
7	Italy	627	2910	212
8	Malaysia	609	132	146
9	Indonesia	523	1060	81
10	Turkey	518	3655	75

.

Figure 6 below displays the countries’ co-authorship network, highlighting the contributions of various countries to technological research in the CCI forecasting field. This map was developed to illustrate the number of documents in which each country has participated.

This analysis revealed a landscape of international contributions to the field, with the United States being the leader in the number of documents produced, with 3086 documents, a staggering total of 20,888 citations, and the highest total link strength of 637, which shows the openness of the US to work with other countries to elevate the research efforts and the significant and far-reaching impact on the global research community. The US was followed by China, which contributed 2121 documents, with 7414 citations across all documents. China had a total link strength of 410, indicating that the country is increasingly becoming a cornerstone of global scholarly conversations in this field.

India came in third place, with 1953 documents, 1003 citations, and a total link strength of 196. The small number of citations compared to the number of documents shows that Indian research is still trying to find the needed grip in the global academic community. Moving to Europe, the United Kingdom’s and Germany’s numbers indicate a great contribution to the research field. The United Kingdom came in fourth place, with 1346 documents, 232 citations, and the second largest total link strength of 528, indicating a well-established research network with the other countries and authors. Germany, with 810 documents, 808 citations, and a total link strength of 253, demonstrated a balanced effort towards the field. Italy was next, with 627 documents, 2910 citations, and a total link strength of 212, signifying that the country’s research work is well-regarded and influential in the field.

Moving towards the Asian part of the world, Australia had 636 documents, 197 citations, and a total link strength of 260. Malaysia had 609 documents, 132 citations, and a total link strength of 146. Malaysia was followed by Indonesia, with 523 documents, 1060 citations, and a total link strength of 81. Turkey showed a stronger impact, with 3655 citations, despite having a smaller document number of 518, which is an indicator of the high quality and relevance of its research contributions.

5.4. Citation Analysis

5.4.1. Academic Sources

The citation analysis started with a review of the most productive sources and journals in the field of cost estimation and CCI studies. The minimum number of documents per source was set to 10 documents and the minimum number of citations was set to 20 citations. The top productive sources are listed in

Table 9. Most productive sources.

№	Source	Documents	Citations	Avg. Pub. Year
1	IOP Conference Series: Materials Science and Engineering	187	656	2019.3
2	International Journal of Forecasting	165	2587	2019.52
3	Applied Sciences (Switzerland)	119	1082	2021.57
4	Lecture Notes in Civil Engineering	112	81	2022.17
5	Journal of Advanced Research in Dynamical and Control Systems	98	120	2018.99
6	Buildings	95	737	2022.33
7	Problems and Perspectives in Management	82	551	2019.56

.

The results show that “IOP Conference Series: Materials Science and Engineering” was the source with the highest number of documents, with 187 publications, receiving 656 citations across all documents. International Journal of Forecasting came in second place, with 165 documents, 2587 citations, and an average publication year of 2019.52, which shows the importance of forecasting and its methodologies being published in the field of construction cost indexes. Next was Applied Sciences (Switzerland), indicating that innovative technologies and applied sciences are vital for the process of construction cost index forecasting, especially given its recent average publication year of 2021.57.

“Lecture Notes in Civil Engineering” had 112 documents, 81 citations, and an average publication year of 2022.17. This was followed by Journal of Advanced Research in Dynamical and Control Systems, which had 98 documents and 120 citations, with an average publication year of 2018.99. Buildings had 95 publications, with a significant citation count of 737 and a recent average publication year of 2022.33, suggesting a growing interest in research related to the construction and management of buildings, which directly impacts cost prediction models and strategies within the industry. Lastly, Problems and Perspectives in Management had 82 documents with 551 citations and an average publication year of 2019.56.

5.4.2. Most Cited Documents

To follow the fast changes in cost estimation and forecasting technologies, it is important to review and study papers and articles that are influential in such developing research fields. This is where citation analysis is beneficial, as it allows one to view the most cited studies, making it possible to determine which publications are reliable and insightful, and which can be considered a source of information in any field of research.

The top 10 most cited documents discuss a range of research topics, factors, and technologies that significantly affect the construction cost and the CCI (see

Table 10. Most cited documents.

№	Author	Document	Year	Citations	Reference
1	Makridakis S.	The forthcoming Artificial Intelligence (AI) revolution: Its impact on society and firms	2017	868	[49]
2	Heckmann I.; Comes T.; Nickel S.	A Critical Review on Supply Chain Risk–Definition, Measure and Modeling	2015	677	[50]
3	Van Hoek R.	Research opportunities for a more resilient post-COVID-19 supply chain	2020	561	[51]
4	Baryannis G.; Validi S.; Dani S.; Antoniou G.	Supply chain risk management and artificial intelligence: state of the art and future research directions	2018	482	[52]
5	Queiroz M.M.; Telles R.; Bonilla S.H.	Blockchain and supply chain management integration: a systematic review of the literature	2020	482	[53]
6	Wangler T.; Loret E.; Reiter L.; Hack N.; Gramazio F.; Kohler M.; Bernhard M.; Dillenburger B.; Buchli J.; Roussel N.; Flatt R.	Digital Concrete: Opportunities and Challenges	2016	476	[54]
7	Galí J.	Monetary Policy, Inflation, and the Business Cycle	2015	329	[55]
8	Akinosho T.; Oyedele L.; Bilal M.; Ajayi O.; Delgado M.; Akinade O.; Ahmed A.	Deep learning in the construction industry: A review of present status and future innovations	2020	243	[56]
9	Fana, M.; Torrejón Pérez, S.; Fernández-Macías E.	Employment impact of Covid-19 crisis: from short-term effects to long-term prospects	2020	174	[57]
10	Rafiei M.; Adeli H.	Novel Machine-Learning Model for Estimating Construction Costs Considering Economic Variables and Indexes	2018	168	[58]

, below).

Makridakis (2017) showed the huge potential of artificial intelligence (AI) in various industries. With 868 citations under his research, he indicated that AI is able to improve predictive accuracy and analyze large sets of data. He also indicated that the AI transformation is a huge and real phenomenon, and will be the controlling system of most industries in the next 20 years, so it is better to manipulate this intelligence to serve our needs than to be flooded by its risks and problems [49]. This offers a great way to enhance CCI calculations, thus improving the precision of the cost estimation and market trend analysis. Heckmann et al.’s (2015) research was cited 677 times. This research provided a detailed analysis of risk and disruption management within supply chains. Adapting different strategies like financial risk assessment and adaptive supply chain performance evaluation will contribute to a more accurate supply chain management, which can help with decreasing disruptions and problems caused by price change risks [50]. This leads to more reliable CCI calculation, especially when it comes to avoiding the economic changes that influence construction costs, like inflation and monetary policy changes. Van Hoek (2020) [51] addressed a very important and tough period, which was the period after COVID-19. The insight from this study, which had 561 citations, highlights the importance of adaptive strategies in maintaining supply chain continuity. Lack of preparation and long response time to disruption risks were two of the many observations that the research addressed as supply chain resilience vandals [51]. Construction experts can effectively forecast and adjust cost indexes to reflect real-time conditions by understanding the impacts of global disruptions.

Back to the importance of AI models, Baryannis et al. (2018) [52] looked into the integration of AI in managing risks within supply chain management. This research, with 482 citations, indicates that the use of AI in predicting risks and uncertainties related to material availability and pricing, which fall under supply chain risk management, is an important task in vital industries, as AI is more capable of providing more advanced predictive and learning capabilities [52]. AI can contribute to better management of potential cost fluctuations, which improves the Construction Cost Index prediction process. Queiroz et al. (2020) [53], with 483 citations, explore the application of blockchain technology and its integration in supply chain management. Transparency and traceability enabled by blockchain are important for improving the provenance of cost data in construction projects [53]. Therefore, this technology integration helps with more accurate CCI calculations, as it ensures that all financial transactions and material costs are accounted for. Wangler et al. (2016) [54] was next in line, with 476 citations. Here, the authors studied the innovative potential of digital concrete, which represents a digital fabrication technology that reduces material waste and increases efficiency and worker safety. These kinds of advancements influence the CCI by lowering construction costs and improving estimation models. Galí (2015) [55] published an impressive book that discussed the impact of macroeconomic factors, including monetary policy, inflation in industries, businesses, and different cost aspects. These factors, as mentioned earlier, hugely affect the CCI calculation and adjusting process, as they reflect the economic environment, thus improving the precision of cost estimations and helping with more informed decision-making in construction management. Akinosho et al. (2020), with 243 citations, provided a complete deep learning model review of the construction industry. This technology can identify the key areas where it can improve project planning and management as well as enhance construction cost prediction by analyzing huge numbers of historical records to reveal trends that traditional methods might overlook [56]. This approach introduces an adaptive management strategy that can be a great asset for large construction projects, where accurate budgeting is a must.

Fana et al. (2020) further underscored the relevance of such an approach by examining the impact of the COVID-19 pandemic on labor markets and broader economic conditions. Their research, which had 174 citations, highlights the short-, mid-, and long-term effects of economic crises on employment in various countries. By analyzing both immediate disruptions and projecting future trends, the study offers valuable insights into how crisis lockdowns and disasters can negatively and harshly impact employment rates, especially in countries with low production, high unemployment, and high poverty rates [57]. Refiei and Adeli (2018), with 168 citations, introduced a cost estimation model that integrates deep Boltzmann machines (DBMs), a machine learning technique, along with the softmax layer, one of the regression models (i.e., BPNN or SVM), and economic indexes as inputs for better cost predictions. This model had far fewer estimation errors than backpropagation neural networks (BPNNs) alone or support vector machines (SVMs) alone, especially under dynamic economic conditions. The model uses factors such as liquidity, wholesale price indexes, and CCIs to provide a responsive tool for accurate cost estimations [58].

6. Discussion

6.1. CCI Forecasting: Methods and Advanced Techniques

Forecasting the Construction Cost Index (CCI) accurately is vital for reliable construction budgets. ARIMA and SARIMA are traditional time-series methods that were widely used, but they often struggle with non-linear changes in the construction industry [12,40]. Both of these models are great for short- and medium-term forecasts. However, ARIMA works well with data that do not have a seasonal pattern, whereas SARIMA is great for data that have a seasonal pattern [40]. Still, SARIMA has problems with sudden and discrete CCI jumps [12,40]. The vector error correction model (VECM) is another time-series model that is effective when there is a long-term relationship between the CCI and other economic indicators like GDP or inflation [13,38]. The model requires careful analysis to ensure that the variables used are related over the long term, which makes this model more accurate but more complex than SARIMA. Support vector machines (SVMs) and neural networks, along with smoothing techniques, have proven to be smart and powerful tools capable of recognizing complex patterns and enhancing prediction accuracy, and are among the best options for projects requiring high accuracy. These models, especially smoothing techniques, are excellent at handling different complex factors that may affect the CCI [10].

Multivariate adaptive regression splines (MARS) and radial basis function neural network (RBFNN) are two other models that can be combined to form a self-adaptive structural radial basis neural network intelligence machine (SSRIM), which is considered a hybrid model that can provide accurate and stable forecasts in complex situations [9]. They are mainly used for projects with high-dimensional data, but they require a lot of experience to be utilized correctly. K-nearest neighbors (KNN) is suitable for less complex projects, as it offers simplicity in data processing, while long short-term memory (LSTM) neural networks are great at capturing both short- and long-term trends, making them suitable for projects with extended timelines.

Choosing a forecasting method depends on the project’s data complexity. For complex situations that need high accuracy, SVM and LSTM are the best models. SARIMA and ARIMA are great for short- or mid-term projects that are simple, whereas SARIMA is great when the data have a seasonal pattern. The VECM is excellent for projects that require an understanding of the short-term dynamics and long-term relationships between the CCI and economic indicators, such as GDP and inflation. K-nearest neighbors (KNN) is suitable for less complex projects. MARS and RBFNN are for complex data that require multiple systems to manipulate. These advanced models are transforming the CCI forecasting process, making it more accurate and reliable.

6.1.1. Identification and Classification of Leading Indicators Influencing the CCI

One of the aims of this study was to define and classify leading indicators and factors affecting the Construction Cost Index (CCI). The effectiveness of CCI forecasting models relies heavily on these indicators for recognizing the variables that impact CCI values. The indicators identified can be grouped into two main categories: macroeconomic indicators, and project-specific indicators.

Macroeconomic indicators are considered broad factors that affect large parts of the construction industry and influence the CCI mostly over the long term. Gross domestic product (GDP) is considered one of the important leading factors, as it has a strong influence on the need for construction services. When GDP grows, the demand for construction services often rises, leading to an increase in both material and labor expenses, impacting the CCI [59]. Another important factor is inflation, as an increase in inflation will certainly lead to an increase in the overall prices and the expenses of construction projects [59]. Interest rate is also considered a vital indicator, as it affects the borrowing cost for the project. This means that project financing will be harder and extra costly, which in turn affects the construction costs and, ultimately, the CCI.

The cost of energy, transportation, and material production is greatly affected by crude oil prices [60], which is considered an important leading indicator affecting CCI forecasting, as was previously indicated in the analysis. The analysis also showcased that monetary policies, which are formed by central banks to affect interest rates and investment levels, can have a huge impact on construction costs. Exchange rates also play an important role in CCI forecasting, as changes in the value of a currency can affect the costs of construction materials, equipment, and energy [61], leading to unknown changes in CCI estimation. Lastly, the unemployment rate is considered to have a huge effect on the CCI, as it can infuse the competition for skilled workers, changing labor wages and thus impacting construction costs [48].

On the other hand, project-specific indicators are tailored for specific construction projects and regions, which can heavily influence the CCI mostly in the short term. These indicators change rapidly and align with specific market situations, which lead to considerable diversity across various projects. Labor availability and wages represent two great indicators that depend on each other, as shortages in skilled labor drive up wages, which in turn increases overall project costs [62]. Similarly, the cost of materials such as cement, steel, and lumber change based on their demand, which can have an immediate and huge impact on project costs. The prices of these materials are often affected by factors such as supply chain disruptions, which is another important factor caused by sudden global events like pandemics, geopolitical issues, or national disasters, leading to material shortages and price volatility, which can have a direct effect on CCI.

The Consumer Price Index (CPI) is another important indicator, as it represents the changes in the prices of goods and services [10]. An increase in the CPI means higher prices for services and materials needed, which leads to an increase in the CCI. Countries with open trade regulations and policies contribute to the stability of the CCI, as the prices are more stable and the material costs are predictable, meaning a noticeable effect on CCI forecasting. Lastly, money supply is considered a huge indicator that affects many industries, especially the construction industry [63]. An increase in money supply can lead to higher inflation, which can increase construction costs and impact the CCI forecasting process, deeming money supply an important leading indicator.

By categorizing the leading indicators and factors of the CCI into macroeconomic and project-specific groups, this study provides a well-organized structure for perceiving the influence of these various factors on CCI predictions. This way of grouping improves the accuracy and robustness of the forecasting models by accounting for both long-term patterns driven by broader economic conditions and short-term changes influenced by project-specific factors. This approach improves insight into the factors influencing CCI changes and reinforces the accuracy of models for predicting construction cost indexes.

6.1.2. Analysis of CCI Forecasting Models Based on Identified Factors

In an earlier section of this research, several methods for forecasting the CCI were examined. This section specifically analyzes the behavior and the ability of these models to adapt and respond to the macroeconomic and project-specific indicators, providing an understanding of their effectiveness in different construction and economic situations.

SARIMA models are great for dealing with seasonal patterns in data that can be found in indicators like material costs or labor availability, which is a part of project-specific factors. On the other hand, SARIMA’s limitations become clear when macroeconomic changes happen suddenly, such as inflation variations, which deems the need for more adaptive models that can handle these influences, especially from external factors, as showcased later in this research. SVM models are effective when it comes to handling complex, non-linear relationships between indicators such as interest rates, exchange rates, and inflation, as these propose unpredictable and interconnected behaviors that affect CCI prediction. The flexibility in SVM’s models to accommodate many indicator combinations makes it great in situations where rapid changes in macroeconomic indicators directly impact construction cost indexes [10,11].

LSTM models are suitable for long-term dependencies when predicting the CCI, especially in markets where GDP growth, money supply, and crude oil prices have a large impact. LSTM allows the storage of historical data over time, which makes the handling of long-term shifts more effective. However, LSTM may be less effective when dealing with short-term disruptions, such as supply chain issues or labor shortages. Next is the VECM, which is effective in handling long-term relationships between factors like the CPI, crude oil prices, and GDP. However, this term captures both short- and long-term trends, making it a reliable choice for CCI forecasting. The VECM can adjust CCI predictions, especially when indicators like GDP or material prices temporarily split from the long-term trends line, ensuring a more accurate forecast.

Self-adaptive structural radial basis neural network intelligence machine (SSRIM) and MARS-based approaches are considered hybrid models that combine multiple forecasting techniques. For instance, SSRIM uses MARS to identify and remove frequent factors, like material cost changes or money supply, whereas networks like RBFNN process the remaining indicators [9]. Hybrid models are perfect for situations where both macroeconomic and project-specific indicators interact, such as simultaneous changes in labor availability, material prices, and GDP, making them useful for predicting the CCI under diverse market conditions. KNN is a simple machine learning model that has been applied especially for short- and mid-term CCI forecasts in environments where indicators like the CPI and crude oil prices fluctuate. However, KNN accuracy increases when additional leading indicators are used, as it relies heavily on training data [14].

Artificial neural network (ANN) models are highly adaptable models that perform well when provided with enough data, as they can handle complex tasks, making them suitable for predicting the CCI with many material indicators such as steel, cement, or bricks, which can change significantly.

6.2. Trends in CCI Forecasting: Key Points

The integration of multiple aspects, such as economic indicators, modeling techniques, risk management methods, and supply chain dynamics, forms up-to-date advancements in Construction Cost Index (CCI) forecasting. Macroeconomic indicators like GDP, interest rates, inflation, crude oil prices, and unemployment are considered the foundations for predicting the CCI. These indicators, as discussed earlier, offer insights into the changes that affect the forecasting process, leaving experts with useful information to prepare for any situation that might affect the overall project budgets. Also, the use of advanced computational techniques, such as machine learning, AI, and deep learning, has improved the accuracy and flexibility of CCI forecasts, as they continuously learn from new project data. Techniques like LSTM, SVM, ARIMA, VECM, and others have shown better efforts to capture fluctuations in different areas and different economic conditions.

Furthermore, hybrid models like SSRIM, which combine multiple forecasting techniques, have been widely used recently to improve CCI prediction accuracy and reliability. These models combine the points of strength of different modeling methods to investigate the complexity of the forecasting data, especially leading indicator data, to generate more robust results for experts. Forecasting models’ adaptability can be improved through the integration of real-time data, as this allows for the accounting of many variables that represent the current market conditions for increased effectiveness of the prediction model.

Another important trend is the importance of risk management in anticipating the impacts of global disruptions, such as pandemics, natural disasters, or geopolitical crises, on construction cost estimation and supply chains. These impacts include huge changes in material prices and other important CCI indicators, making accurate forecasting more challenging. Many studies, especially the ones discussed in this research, have addressed the role of successful supply chain management in avoiding risks and sudden market fluctuations.

Blockchain technology is gaining recognition as a valuable tool for enhancing data transparency in multiple subjects, with the supply chain being at the top of them. Blockchain provides a decentralized platform that is perfect for material tracking and financial transactions, ensuring reliability in supply chain management and providing a reliable foundation for better data management. However, in addition to its role in improving data management and cost prediction, blockchain technology improves sustainability in construction by enhancing resource management and reducing environmental impact [64]. It also can minimize inefficiencies in procurement, which can help with preventing over-ordering and excess inventory, leading to a smarter approach to using resources and reducing the harmful carbon footprint of construction projects. Blockchain can help promote sustainability within the construction supply chain by encouraging transparency and accountability [64]. Similarly, AI and machine learning can help improve resource allocation by accurately predicting demand changes. This allows construction firms to avoid overconsumption of raw materials, boosting sustainability standards and minimizing waste, which also aligns with the global efforts to reduce construction waste and improve efficiency [65]. This clearly shows how smart technologies like blockchain, advanced models like AI, and efficient supply chain management can lead to more responsible resource management and a reduced environmental impact in construction projects, adhering to sustainability goals such as those set by LEED and other green building certifications.

6.3. Detailed Bibliometric Analysis

6.3.1. Co-Occurrence Analysis Insights

This analysis offered a detailed visualization of the keywords and their designated clusters within the comprehensive CCI literature. The main keywords included “forecasting”, “supply chains”, “construction costs”, and “artificial intelligence”, and the combination of these keywords together indicates the importance of the predictive process and technological integration in construction cost management. These interconnected areas point to future research integrating AI and supply chain management for better cost index forecasting and project planning.

The clustering of the keywords was based on their connection in various research areas, forming a total of five main clusters, which were (1) supply chain management, (2) economic indicators, (3) forecasting and data science, (4) construction costs and project management, and (5) technology and AI integration. These clusters highlight the current research trends and identify areas that are likely to gain importance nowadays and in the future.

6.3.2. Co-Authorship Analysis Insights

This analysis highlights the significant efforts and collaborations among researchers. Key contributors like Sawik T. and Ozili P.K. have shown high productivity and great influence within the field, which helped with the development of CCI prediction models and addressing its affecting factors and methods. The geological distribution of research efforts was also highlighted in this analysis. It indicated that the United States and China lead the publication production and citation impact in this field, which reflects their dominance and their pivotal role in cultivating CCI research, driven by their huge construction industries and academic resources.

6.3.3. Citation Analysis Insights

This analysis provided insights into the most influential documents and sources within the CCI research field. Foundational studies such as Makridakis (2017) [49], Heckmann et al. (2015) [50], and Rafiei and Adeli (2018) [58] have been widely cited for their important influence on AI, supply chain, and machine learning applications, which can be used for future investigation in construction cost estimation. Journals and sources like International Journal of Forecasting, Applied Sciences, and Buildings are highly respected platforms for their huge and effective impact on the research industry, especially in the Construction Cost Index forecasting and estimation field.

6.4. Implications for Future Research and Studies

6.4.1. Advancing CCI Forecasting Methods

This research highlighted several critical areas for further exploration. The use of advanced models and multiple smart approaches, as well as identifying the trends and affecting factors, has pushed the field, but a lot of opportunities remain to enhance the accuracy and adaptability of CCI forecasting and calculation.

Effective forecasting models used for time-series data, like SARIMA, have problems with discrete jumps in the CCI [12]. Future research could explore methods or enhanced alternatives to capture these sudden jumps by studying different hybrid models that combine multiple techniques to improve model response to these sudden changes. Also, it was mentioned earlier that inflation, GDP, and monetary policies are considered economic indicators that play a critical role in forecasting the CCI [6]. Therefore, a future research opportunity lies in developing dynamic models that can combine these factors in real time through the use of AI and machine learning to produce accurate and more responsive CCI forecasts. Similarly, simulations can be created that can predict different economic policy changes based on current factors to account for the changes before the CCI forecasting process even begins.

Furthermore, advanced modeling techniques like machine learning or others can incorporate other factors like supplier flexibility, multiple sourcing options, and the time of response, which represent some of the main causes of supply chain disruptions [51]. This can help with predicting cost fluctuation, which can help with more adaptive and reliable CCI forecasting without the excessive focus on cost-related indicators. On the other hand, the observed regional differences in markets suggest a need for forecasting models that are unique for each of these markets. Previous studies have completed the work of gathering the prediction indicators for China [18] and Malaysia [44], but the difference between the amount of data gathered and the construction industry situation can affect the type of model used. Future research can focus on developing region-specific models that account for the local material and labor costs as well as economic conditions. In addition, creating models for specific types of construction, such as residential or commercial, would help with more accurate CCI forecasting and cost estimation, showing the true dynamics of each construction sector.

Lastly, future research could address the idea of using blockchain to improve CCI prediction through the enhancement of the data used for the forecasting process, as the amount of research on this topic is close to none.

6.4.2. CCI Forecasting and Sustainable Construction

The demand for sustainability in construction is increasing, meaning new areas and opportunities for further research, especially when it comes to the intersection of CCI forecasting and sustainable building practices. Future studies could investigate the incorporation of environmental indicators, such as carbon emissions, waste production, or reduction and energy consumption metrics, into the current CCI forecasting models. This can help construction companies gain valuable insights on the environmental impacts of their decisions in addition to the project cost estimations.

Advanced technologies such as blockchain can be further studied to address how their integration can enable the tracking of environmental factors like the carbon footprint associated with each construction material alongside traditional supply chain data, providing real-time data for more informed sustainable resource management decisions. Also, machine learning can be further utilized to integrate both cost and sustainability indicators, creating hybrid models that balance financial goals with environmental responsibility. These hybrid models can greatly contribute to advancing sustainable construction techniques by improving how resources are used, lowering waste, and aiming for long-lasting sustainability results.

Finally, future research could also explore incorporating life cycle assessments (LCAs) into CCI forecasting models, helping the construction industry address the environmental impact of different construction methods, indicators, and materials and move towards sustainable building practices and less environmental degradation.

7. CCI Prediction: Case Study

The seasonal autoregressive integrated moving average (SARIMA) model was applied to forecast the cost index for highway construction in Michigan. The process started with the collection of historical data for the CCI for every quarter, starting from the first quarter of 2010 until the fourth quarter of 2019 [66], to be used in the implementation process along with the forecasting model (see

Table 11. Historical CCI data.

QUARTER	CCI	QUARTER	CCI	QUARTER	CCI
20101	1	20133	1.0906	20171	1.079
20102	0.9847	20134	1.0571	20172	1.0749
20103	0.9893	20141	1.0797	20173	1.2304
20104	1.0046	20142	1.0944	20174	1.1861
20111	0.974	20143	1.2051	20181	1.2076
20112	1.063	20144	1.1624	20182	1.351
20113	1.0819	20151	1.1074	20183	1.3812
20114	1.0478	20152	1.1543	20184	1.3352
20121	1.0444	20153	1.2843	20191	1.3227
20122	1.0695	20154	1.1359	20192	1.4672
20123	1.0213	20161	1.1428	20193	1.4851
20124	1.0522	20162	1.1791	20194	1.4956
20131	1.0531	20163	1.2197
20132	1.0961	20164	1.0901

). This dataset served as a foundation for the implementation of the SARIMA model.

The historical CCI data were then trained and tested. The results of the analysis showed that the seasonal ARIMA (AR = 0, I = 0, MA = 0) (seasonal AR = 1, seasonal I = 1, seasonal MA = 1) model outperformed the ARIMA model in forecasting the quarterly CCI [66]. The parameters of this model consisted of two parts: the non-seasonal component and the seasonal component. Non-Seasonal Component: (1) AR (autoregressive) = 0, which means that the model does not rely on past values to predict future values; (2) I (integrated) = 0, which means that the data are already stationary; and (3) MA (moving average) = 0, meaning the model does not take errors in previous forecasts into account. On the other hand, the seasonal component consists of (1) seasonal AR (autoregressive) = 1, which means that the model accounts for the relationship between past values in previous quarters; (2) seasonal I (integrated) = 1, meaning that the model applies one differencing operation to account for seasonal trends in the data; and (3) seasonal MA (moving average) = 1, which means that the model accounts for forecast errors that repeat in a seasonal pattern.

The implementation of the SARIMA model provided important insights into the reliability of the predictions revealed. The prediction model was run for a six-quarter period, starting from the third quarter of 2018 until the fourth quarter of 2019. The predicted values were compared to actual CCI values for the specified quarters, as shown in

Table 12. Actual CCI values vs. predicted values.

Quarter	Actual	Predicted
20183	1.3812	1.4735
20184	1.3352	1.3687
20191	1.3227	1.3711
20192	1.4672	1.4441
20193	1.4851	1.5890
20194	1.4956	1.4859

below.

Figure 7 represents the relationship between the actual and predicted values over the six specified quarters. The red line represents the actual CCI values from the third quarter of 2018 until the fourth quarter of 2019, showcasing the fluctuations in the Construction Cost Index. The figure also shows the predicted values represented by the blue line, presenting the overall prediction trend compared with the actual trend for the same period.

To evaluate the effectiveness of the forecasting model, the predicted CCI values were compared with the historical values using several performance metrics recently mentioned in this research, which are (1) mean absolute error (MAE), (2) mean squared error (MSE), and (3) mean absolute percentage error (MAPE). Their values are shown in

Table 13. Performance metrics for the prediction model.

	MSE	MAE	MAPE
SARIMA	0.0039	0.0518	3.68%

. The MAE indicates that the average difference between the predicted values and the actual ones was 0.0518 index points, marking a high level of accuracy. The MSE with a value of 0.0039 highlights a relatively minor overall deviation from the actual numbers, while the MAPE of 3.68% represents the model’s average error in relation to the actual values.

In summary, the performance metrics indicate that the utilization of SARIMA as a prediction model in the forecasting of the CCI showcases its capability to capture seasonal trends. The metrics also suggest that while the SARIMA serves as a reliable baseline for forecasting, it still has the potential for improvement in prediction accuracy.

These findings highlight the importance of addressing the underlying complexities and the inability to model dynamic economic relationships, as well as considering the potential external influences that may affect the prediction’s reliability and accuracy. Moreover, industry experts can apply similar models in other construction sectors where seasonal patterns are important. Future studies must modify the model by refining its parameters and integrating additional relevant data to increase its forecasting accuracy. By doing so, stakeholders in the construction industry can enhance their forecasting skills, leading to more informed decisions and improved project, cost, and resource management.

8. Conclusions

This paper advances the understanding of Construction Cost Index (CCI) forecasting models by focusing on their prediction accuracy and reliability in project management and construction cost control. The study highlights the complexity of predicting the CCI due to various identified macroeconomic and project-specific indicators, such as inflation, GDP, and supply chain disruptions. The bibliometric analysis showed the emerging trends in the CCI literature, including the increasing focus on modeling techniques, economic indicators, and the key challenges in construction cost estimation. Furthermore, the co-authorship, co-occurrence, and citation analyses helped identify keywords, authors, documents, and countries that are driving this field and contributing to its development.

A case study using the SARIMA model demonstrated its practical application in forecasting the CCI, showcasing its ability to capture seasonal trends and highlighting the areas that need further improvement. This application deems the importance of refining forecasting models to improve their effectiveness in real-world construction scenarios.

Moreover, the research suggests that sustainable construction practices should be integrated as part of future prediction models, especially when incorporating factors like resource efficiency and environmental impact in the prediction process. Advanced technologies like AI and blockchain are recognized for their potential to enhance transparency, sustainability, and resource management in the construction industry, contributing to more responsible project outcomes.

Looking forward, future research should focus on further development and the integration of adaptive forecasting models that can react more dynamically to market changes and external influences and shocks. The combination of advanced technologies with flexible and responsive models will empower the construction industry to manage both cost volatility and sustainability goals more effectively, promoting a more resilient and future-oriented approach to cost management in the construction industry.