AbstractDetermining a reasonable project duration is one of the most critical activities required by project owner agencies for successful project letting and delivery. Most owner agencies, specifically in the highway sector, mainly rely on schedulers’ judgment and experience in determining the sequence of construction activities to estimate the required amount of time of a project. A vast amount of historical project performance data available in owner agencies’ databases provide rich and reliable resources that can significantly improve the current process to produce a consistent and repeatable quality of construction logic determination. This study proposes a novel data-driven process model utilizing pattern mining, statistical analysis, and network analysis techniques that can detect pairwise logical relationships among construction activities (e.g., Start-Start and Finish-Start) and develop knowledge networks of as-built construction sequence patterns to improve the scheduling process. Three algorithms are proposed to apply the knowledge networks to sequencing a new project: finding immediate predecessors and successors of an activity or ordering a given set of activities. Ten years of historical project data obtained from a state department of transportation were used in this research. A case study reveals that the process model developed in this study can successfully build the most reasonable construction sequences of a highway project, which can significantly improve the scheduling and contract time determination process.