Developing a machine learning model optimized with the jellyfish search algorithm for predicting labor productivity on construction sites

Abstract

Construction activities are significantly dependent on labor productivity due to its direct impact on economic efficiency and project progress. Therefore, enhancing labor productivity on construction sites remains a top priority for businesses and construction management experts. This study presents a comparative evaluation of the performance of various machine learning models, including four individual models: ANN, SVR, LR, CART, and three ensemble models: Voting, Bagging, and Stacking. The results demonstrate that the Bagging-ANN ensemble model yields the highest efficiency. The model's parameters are optimized using the Jellyfish Search algorithm to improve its performance. The final results are compared with literature, revealing the superior performance of the JS-Bagging-ANN model.

Keywords: Jellyfish Search; labor productivity; machine learning models, optimization; prediction system.