Identifying Structural Drivers of Crime Using Statistical Inference and Machine Learning Models

This study identifies structural socio-economic correlates of crime investigation activity in Kazakhstan and examines how interpretable composite indicators relate to it. The dependent variable is the annual volume of pre-trial investigations recorded in the Unified Register of Pre-Trial Investigations over 2015–2025. It is modeled against 70 explanatory factors that span socio-economic, demographic, institutional, legal, healthcare, education, agriculture, and digitalization domains. The analysis combines Pearson correlation, one-way analysis of variance, simple and multivariate linear regression, and regularized ensemble machine learning (Gradient Boosting, Random Forest, Ridge, Support Vector Machine). A category-based dimensionality reduction then aggregates the 70 factors into 8, and subsequently 10, standardized composite indices. Because only about 10 aggregated annual observations are available, findings are reported in associational rather than causal terms. The strongly negative cross-validated R2 of the ensemble models is flagged as direct evidence of overfitting in this small-n, high-dimensional setting. Digitalization, education, poverty, and healthcare indicators emerge as the leading correlates of crime. A parsimonious regression on the composite indices accounts for 97.6% of insample crime variance while improving interpretability and stability. Interpretable index-based models therefore appear better suited than high-capacity learners for policy-relevant inference in this setting. Received: 26 January 2026 | Revised: 6 August 2026 | Accepted: 21 September 2026 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement Data are available from the corresponding author upon reasonable request. Author Contribution Statement Alibek Barlybayev: Conceptualization, Writing – original draft, Writing – review & editing, Supervision, Project administration, Funding acquisition. Mira Kaldarova: Methodology, Validation, Formal analysis, Investigation, Resources, Data curation. Rakhat Yerezhepov: Software, Validation, Resources, Data curation, Visualization. Nurlan Botabayev: Validation, Formal analysis. Madi Tugunbayev: Methodology, Resources, Supervision, Funding acquisition. Gulmira Shakmetova: Writing – review & editing, Software, Validation, Data curation, Visualization.