A Sequential Pattern Mining Driven Framework for Developing Construction Logic Knowledge Bases
One vital task of a project's owner is to determine a reliable and reasonable construction time for the project. A U.S. highway agency typically uses the bar chart or critical path method for estimating project duration, which requires the determination of construction logic. The current practice of activity sequencing is challenging, time-consuming, and heavily dependent upon the agency schedulers' knowledge and experience. Several agencies have developed templates of repetitive projects based on expert inputs to save time and support schedulers in sequencing a new project. However, these templates are deterministic, dependent on expert judgments, and get outdated quickly. This study aims to enhance the current practice by proposing a data-driven approach that leverages the readily available daily work report data of past projects to develop a knowledge base of construction sequence patterns. With a novel application of sequential pattern mining, the proposed framework allows for the determination of common sequential patterns among work items and proposed domain measures such as the confidence level of applying a pattern for future projects under different project conditions. The framework also allows for the extraction of only relevant sequential patterns for future construction time estimation.
Le, Chau; Shrestha, Krishna J.; Jeong, H. D.; and Damnjanovic, Ivan. 2021. A Sequential Pattern Mining Driven Framework for Developing Construction Logic Knowledge Bases. Automation in Construction. Vol.121 https://doi.org/10.1016/j.autcon.2020.103439 ISSN: 0926-5805