Data-Driven Process Mining Framework for Risk Management in Construction Projects

Abstract

Abstract Construction Projects are exposed to numerous risks due to their complex and uncertain nature, threatening the realization of the project objectives. However, Risk Management (RM) is a less efficient realm in the industry than other knowledge areas given the manual and time-consuming nature of its processes and reliance on experience-based subjective judgments. This research proposes a Process Mining-based framework for detecting, monitoring, and analysing risks, improving the RM processes using evidence-based event logs, such as Risk Registers and Change-Logs within previous projects’ documents. Process Mining (PM) is a data-driven methodology, well established in other industries, that benefits from Artificial Intelligence(AI) to identify trends and complex patterns among event logs. It performs well while intaking large amounts of data and predicting future outputs based on historical data. Therefore, this research proposes a Bayesian Network (BN)-based Process Mining framework for graphical representation of the RM processes, intaking the conditional dependence structure between Risk variables, and continuous and automated risk identification and management. A systematic literature review on RM, PM, and AI forms the framework theoretical basis and delineates the integration areas for practical implementation. The proposed framework is applied to a small database of 20 projects as the case study, the scope of which can be tailored to the enterprise requirements. It contributes to creating a holistic theoretical foundation and practical workflow applicable to construction projects and filling the knowledge gap in inefficient and discrete conventional RM methods, which ignore the interdependencies between risk variables and assess each risk isolated.