A Methodology and Tool for the Predictive Analysis of Cost Growth in Construction Projects

The difference between the actual cost and the budgeted cost is high in construction projects. Due to their complex nature, it can be challenging to understand the true impact of each data field on the overall execution and completion of the project. There is therefore a need for an interactive tool to allow for predictive analysis based on different scenarios and variations of significant data fields. This paper presents the analytical techniques and tools that provide an interactive predictive analysis of large-scale projects with different features, and outlines how the tool's design can help practitioners generate insight and assist them in making strategic decisions for their company. The designed tool uses five conventional machine learning approaches and deep learning algorithms to predict project cost growth. The tool uses two groups of features, 16 important features selected by the previous study and 29 features given by the domain expert, which are prior to the beginning of the construction phase. Different combinations of factors and different models can be selected for prediction. The tool can help practitioners know about the cost growth of a new project and its corresponding R2 score and RMSE at different project phases. The number of samples is limited to 139 projects; therefore, having more data can improve performance.