Abstract
Targeted police stops are frequently carried out by police in response to real-world needs. The effectiveness of various purpose-driven police stop tactics on crime prevention and control varies. However, existing research has neither identified the associated factors of police stops nor explored their impact on crime with different factors. Therefore, this study focuses on the main urban areas of megacities along the southeast coast of China. The space is partitioned using hierarchical clustering after applying the XGBoost and SHAP algorithms to determine the factors related to police stops. Lastly, this study explores the causal effects of police stops with different associated factors on crime, using causal forests within double machine learning. There are three conclusions. First, there is a strong correlation between police stops and four variables: alarm, visiting population, criminal, and government agencies. Second, by clustering based on different associated factors of police stops, existing police stops can be classified into five categories according to their purposes: (i) composite stops positively associated with “Alarm, Visiting Population, Criminals” (AVC-CPS); (ii) composite stops positively associated with “Alarm, Visiting Population, Bus Station” (AVB-CPS); (iii) random stops with no significant positive association (NA-RPS); (iv) single police stops positively associated with “Alarm” (A-SPS); and (v) single stops positively associated with “Visiting Population” (V-SPS).
— Read on www.nature.com/articles/s41599-025-05355-0