Construction Quality Risk Assessment Models for Justifying Inherent Defects Insurance: Quantified Evidence from Big Data in Court Cases

Abstract

An accurate risk assessment of construction quality helps in project selection and management for inherent defects insurance (IDI). However, existing quality risk-assessment studies often relied heavily on expert opinions and always paid attention to surrounding environment of construction projects, lacking a focus on the inherent quality risks to the construction itself. This research aims to develop an accurate construction quality risk assessment model (CQ-RAM) through assessing risk quantitatively based on legal judgments related to construction quality problems and repairing costs confirmed by courts. Three research stages were designed. First, a preliminary indicator system for construction quality risk assessment for IDI problem terms was established based on existing literature and practical project quality records. Second, by referring to court judgments (sample size: N=32,368), the consequences of each quality risk indicator were quantified according to repair costs through graph topology analysis. Third, some CQ-RAMs were developed and tested by comparing the evaluation results using random forest (RF) model, support vector machine (SVM) model, and artificial neural network (ANN) model, respectively. The study found the RF-based CQ-RAM model was the best predictive system in terms of its effectiveness and accuracy for quantifying the risk. The finding of the study provides a new basis for risk assessment for construction quality, especially for insurance companies in deciding IDI underwriting.