Developing a Model to Estimate Plastering Labor Productivity using Artificial Intelligence

Abstract

    Underestimating productivity carefully causes the project to be delayed and the cost to be increased; therefore, estimating productivity is of great importance in all aspects of construction. The human factor generally outweighs other factors that affect productivity since a large percentage of jobs depend on it. Plastering is one of the most important internal and external cladding works for all types of projects. There is hardly any building or facility in Syria devoid of plastering works. Therefore, this research aims to develop an Artificial Neural Network (ANN), which does not depend on personal experience and connects the factors that affect labor productivity in plastering with each other. The factors that affect productivity are determined through the previous studies and by collecting opinions of project managers, site managers, contractors, and through the data acquired. Eight influential factors are used to predict labor productivity in plastering: location of plastering, method of plastering, number of plastering patches, number of work teams, age, work team experience, weather, and floor height. The network is trained on several architectural designs, and the design with the least mistakes and the highest correlation coefficient is chosen as the best architectural design. The optimal network is composed of eight inputs to the network, a hidden layer with 10 nodes, and the output layer is related to labor productivity. The optimal trained network is tested on new samples. The test results show the preciseness of the suggested model in prediction with a correlation coefficient of R=98.12%. Finally, a software interface is developed that provides labor productivity in plastering using the abovementioned network.