Work rate-based indicators for improving project performance regression models

Performance analysis is a critical component of project monitoring that provides the basis for evaluating cost and duration estimates at completion. Although Earned Value Analysis (EVA) techniques such as Earned Value Management (EVM) and Earned Schedule (ES) are widely used, they only focus on cost and time features and overlook the relationship between cost and schedule performances. Previous works have provided additional indicators and demonstrated the implications of work rate variations on project performance, but have not integrated the two streams. Thus, this study aims to improve project monitoring by developing and validating three novel indicators for increasing the accuracy and precision of regression-based performance forecasting models. The indicators quantify the project's relative phase, work rate, and work rate acceleration. Validation was performed using 60 real construction project data by comparing the performance of multilayer perceptron (MLP) neural network models with and without these indicators in cost and duration forecasting. Then, the MLPs were benchmarked against EVM and ES models on the same dataset. Results show that including work rate-based indicators in the MLPs improves their accuracy and precision and that MLPs can outperform EVA techniques.