Construction productivity forecasting modelling

characterization and comparative critical analysis on a case study

Abstract

The productivity rates used as references by construction companies are generally obtained empirically, either through databases of previous projects or based on reference indices from budgeting manuals. However, the use of average productivity indicators represents an overly simplistic approach considering the current need for a deeper understanding of construction activities, given the large number of content and context factors that can influence services. An alternative for predicting productivity lies in forecasting models, which are systematic approaches used to develop mathematical or computational representations that describe the reality of a system, process, or phenomenon. Thus, this study aims to apply and compare four different modeling techniques for productivity forecasting, including two statistical models and two artificial intelligence models. The productivity forecasting was carried out based on nine content and context input factors deemed significant for concrete formwork execution services. The different models employed were evaluated. The results demonstrate that it is not always possible to find the best accuracy parameters within a single model.