Time series clustering reveals three evolution patterns of duty-related crimes in construction engineering

Background Duty-related crimes in the construction engineering sector are highly prevalent and cause severe harm, undermining infrastructure quality and urban resilience. Methods Based on 93 court verdicts of duty-related crimes from four provinces in China between 2021 and 2023 retrieved from China Judgments Online, this study extracts time series data on the amounts involved and relationship intensity. The Piecewise Aggregate Approximation (PAA) method is employed for dimensionality reduction, followed by K-means clustering with Shape-Based Distance (SBD) to analyze the evolution patterns of criminal behavior. Results The results indicate that the evolution patterns of the amounts involved can be categorized into three types: “stable-conservative type”, “high-level fluctuating type”, and “gradual-stage type”. The evolution patterns of offender relationships can be classified into three types: “hill-shaped fluctuating type”, “low-level fluctuating type”, and “high-level fluctuating type”. The study also finds that all three patterns exhibit convergent characteristics in the later stages of criminal activities, and the distribution of amounts involved shows a “50,000-yuan cycle” pattern. Behavioural interpretations grounded in middle-range theories, such as prospect theory, personality traits, rational choice, bounded rationality, social exchange theory, and impression management, are provided to explain the observed dynamics. Conclusion Understanding these evolution patterns is essential for enhancing urban infrastructure resilience, as corruption diverts resources from quality construction and maintenance. These findings contribute to urban resilience by identifying targeted intervention points along the criminal trajectory. Targeted governance strategies are proposed to provide references for precise prevention and control of duty-related crimes in construction engineering projects.