A Two-Stage Machine Learning-Based Heuristic Algorithm for Buffer Management and Project Scheduling Optimization

One of the main problems that project managers face is that in most cases the projects won’t be completed according to predetermined schedules and therefore prolonged delays and losses occur during the project implementation phase. This study aims to propose a predictive buffer management algorithm (PBMA) based on machine learning technology to predict project buffer size and control project buffer consumption in construction projects to be not more than predetermined schedules and consequently prevent delays in projects’ completion times. The proposed machine-learning-based heuristic algorithm falls into the category of supervised machine learning algorithms and consists of two main stages. In the first stage, the project critical chain is identified and the appropriate project buffer size is specified. In the second stage, the consumption of the project buffer is monitored and controlled in the project implementation stage. To evaluate the performance of the proposed PBMA, it is coded in MATLAB software and implemented using the data taken from a hypothetical project. The results show that the use of the proposed PBMA improves the productivity of projects, and therefore the projects can be completed according to predetermined schedules. The resulted values for the longest path of the project’s critical activities, the duration of the project’s critical chain, the buffer duration of the project, and the duration of the project plan are respectively 60, 30, 15, and 45 days. The proposed PBMA can be applied to a variety of projects.