Scholarlink

<p>This study examines the relationship between dominant land use patterns and crime occurrence using a novel mesh-based geospatial approach integrated with machine learning techniques in Chandgaon Thana, Chattogram, Bangladesh. A 30&#xa0;m × 30&#xa0;m mesh grid was employed to analyze crime exposure across seven land use categories using 2024–2025 crime data and predict 2026 crime probabilities. Chi-square analysis revealed statistically significant associations between land use and crime (χ<sup>2</sup> = 726.9 in 2024; χ<sup>2</sup> = 1066.33 in 2025, <i>p</i> &lt; 0.001), with Cramér's V coefficients indicating strengthening relationships (0.27 in 2024; 0.31 in 2025). Major roads exhibited the highest crime exposure (33.30% in 2025), followed by playground areas (23.63%), while educational and religious facilities showed minimal exposure. Spatial analysis demonstrated that 45.26% of the study area faced high to very high crime risk, with criminal activities concentrating along transportation corridors and mixed settlement zones. Random Forest classification achieved 95.2% accuracy with AUC of 0.84, successfully predicting spatial crime probabilities for 2026. The bimodal probability distribution revealed distinct low-risk (24.64% of area) and high-risk (45.26% of area) zones, indicating significant spatial polarization of crime vulnerability. These findings provide evidence-based insights for targeted crime prevention strategies and urban planning interventions in rapidly urbanizing contexts.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Assessing crime exposure and predictability from dominant land use patterns using an integrated mesh-based geospatial and machine learning approach

  • 期刊论文

  • 发表日期：

    2026年4月9日

查看全文
  • Sajib Sarker,
  • Sajid Ibne Alam,
  • Montaha Tasnim Chowdhury,
  • Najiba Maliyat Hossain,
  • Ariba Rahman

摘要

This study examines the relationship between dominant land use patterns and crime occurrence using a novel mesh-based geospatial approach integrated with machine learning techniques in Chandgaon Thana, Chattogram, Bangladesh. A 30 m × 30 m mesh grid was employed to analyze crime exposure across seven land use categories using 2024–2025 crime data and predict 2026 crime probabilities. Chi-square analysis revealed statistically significant associations between land use and crime (χ2 = 726.9 in 2024; χ2 = 1066.33 in 2025, p < 0.001), with Cramér's V coefficients indicating strengthening relationships (0.27 in 2024; 0.31 in 2025). Major roads exhibited the highest crime exposure (33.30% in 2025), followed by playground areas (23.63%), while educational and religious facilities showed minimal exposure. Spatial analysis demonstrated that 45.26% of the study area faced high to very high crime risk, with criminal activities concentrating along transportation corridors and mixed settlement zones. Random Forest classification achieved 95.2% accuracy with AUC of 0.84, successfully predicting spatial crime probabilities for 2026. The bimodal probability distribution revealed distinct low-risk (24.64% of area) and high-risk (45.26% of area) zones, indicating significant spatial polarization of crime vulnerability. These findings provide evidence-based insights for targeted crime prevention strategies and urban planning interventions in rapidly urbanizing contexts.