Machine Learning Framework to Predict Last Planner System Performance Metrics

Despite numerous attempts toward enhancing performance, the construction industry is still behind in this term. Various technological advancements are at the disposal of construction researchers and practitioners to address this issue. Machine learning techniques are one example of technologies that have become readily accessible to general users thanks to the efforts of researchers in construction planning and control and other fields. Accordingly, the performance in the construction industry may be improved by employing machine learning techniques for developing performance indicators to forecast possible issues or take corrective measures proactively. While some studies have applied machine learning in various aspects of lean construction, no research has yet employed machine learning to predict specific performance metrics. This study, therefore, aims at developing a framework to predict the Last Planner System (LPS) metrics. This will allow optimizing performance on a near real-time basis. A framework for implementing such an approach toward project control is presented, and several predictive models are created, compared, and refined. This approach opens an entirely new avenue for applications of machine learning and data mining techniques in lean construction project planning and control.