A Risk Index Model for Construction Projects Based on Multi-Source Data

Abstract

With the increasing complexity of construction projects and the accelerating pace of urbanization, safety management on construction sites faces growing challenges. To address the limitations of traditional safety management approaches in terms of timeliness, system integration, and intelligence, this paper proposes a novel risk index model for construction projects based on multi-source data. The model categorizes safety risks into five key dimensions: worker, equipment, work activities, subcontractors, and civilization & environment. Over 800 refined evaluation indicators are extracted across these dimensions. Using the Analytic Hierarchy Process (AHP), weights are assigned to each indicator to develop a quantifiable and traceable safety evaluation method. By integrating perception technologies such as AI cameras, sensors, and drones with a multimodal large-model analysis engine, the system enables dynamic risk behavior identification, indicator matching, and automated scoring. The risk scores for each category are aggregated via a MaaS (Management-as-a-Service) platform to generate an overall safety index for the site, supporting functions such as early warning, closed-loop rectification, and re-evaluation. The proposed model establishes a full-cycle intelligent management process covering perception, analysis, evaluation, warning, rectification, and feedback. Pilot implementations demonstrate significant improvements in risk detection rate, response speed, and rectification efficiency, indicating strong engineering applicability and potential for broader adoption. This research offers theoretical support and practical guidance for achieving intrinsic safety and building intelligent construction sites in the industry.