Predicting construction productivity with machine learning approaches

Machine learning (ML) is a purpose technology already starting to transform the global economy and has the potential to transform the construction industry with the use of data-driven solutions to improve the way projects are delivered. Unrealistic productivity predictions cause increased delivery cost and time. This study shows the application of supervised ML algorithms on a database including 1,977 productivity measures that were used to train, test, and validate the approach. Deep neural network (DNN), k-nearest neighbours (KNN), support vector machine (SVM), logistic regression, and Bayesian networks are used for predicting productivity by using a subjective measure (compatibility of personality), together with external and site conditions and other workforce characteristics. A case study of a masonry project is discussed to analyse and predict task productivity.