Artificial Intelligence Framework for Developing a Critical Path Schedule Using Historical Daily Work Report Data

A considerable amount of literature has employed data mining and machine learning (DM & ML) algorithms such as artificial neural networks (ANNs) to predict the duration of a construction project. However, these studies were mostly applied in the planning phase. Highway agencies use the critical path method (CPM) as one of the primary methods for estimating construction time. In contrary with DM & ML prediction models, CPM allows users to leverage detailed information available at the end of the design in estimating construction time and adjust the time if necessary based on the effects of various constraints and influential factors. CPM, however, has limitations, and one of them is its dependency on the expertise of schedulers to produce a reliable estimate. CPM also requires a significant amount of time and efforts from the schedulers to prepare the forecast. With the aims of mitigating those limitations and supporting highway agencies in establishing construction time using CPM, this study proposes a framework that leverages historical digital daily work report (DWR) data to automatically create draft CPM schedules for new projects by exploiting the combined strengths of CPM and DM & ML algorithms. The proposed framework uses activity-level data for estimating construction time such as construction activities, their corresponding quantities, and their start and finish times that are available in DWR data. As a result, the framework can estimate activity production rates using DM & ML models and determine activity relationships using Sequential Pattern Mining algorithms. DWR data collected from a highway agency were used to demonstrate the main components of the framework.