Scholarlink

<p>Crime prediction has become an important challenge in modern cities wanting to enhance public safety and law enforcement. Analyzing crime patterns provides us a valuable insight on criminal behaviour and supports the efforts made for crime prevention. This study tackles the challenge of accurately classifying and forecasting criminal activities using machine learning methods. By examining spatial and time-based crime patterns, this study helps in identifying high-risk zones and predict future incidents based on key attributes such as location, time, seasonal crimes and other crime categories. Utilizing the publicly available Chicago crime dataset from Kaggle that includes a huge amount of city-level incident records, the study develops an integrated prediction framework combining Support Vector Machine (SVM) for crime classification and XGBoost for predictive analysis. To handle data imbalance, the Synthetic Minority Over-sampling Technique (SMOTE) is applied, improving the model’s ability to learn minority crime classes effectively. Experimental evaluation shows that XGBoost achieves a high prediction accuracy of 97.3%, significantly outperforming SVM, which attains 53.2% accuracy. The results confirm that the combined use of SVM and XGBoost provides a reliable and scalable solution for crime prediction.</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

A Machine Learning Approach for Crime Classification and Prediction

  • 期刊论文

  • 发表日期：

    2026年1月7日

查看全文
  • Sanjit Kumar Dash,
  • Surajit Mohanty,
  • Saswat Choudhury,
  • Aanchal Mohanty,
  • Pulak Ranjan Mohanty,
  • Biranchi Kumar Nayak

摘要

Crime prediction has become an important challenge in modern cities wanting to enhance public safety and law enforcement. Analyzing crime patterns provides us a valuable insight on criminal behaviour and supports the efforts made for crime prevention. This study tackles the challenge of accurately classifying and forecasting criminal activities using machine learning methods. By examining spatial and time-based crime patterns, this study helps in identifying high-risk zones and predict future incidents based on key attributes such as location, time, seasonal crimes and other crime categories. Utilizing the publicly available Chicago crime dataset from Kaggle that includes a huge amount of city-level incident records, the study develops an integrated prediction framework combining Support Vector Machine (SVM) for crime classification and XGBoost for predictive analysis. To handle data imbalance, the Synthetic Minority Over-sampling Technique (SMOTE) is applied, improving the model’s ability to learn minority crime classes effectively. Experimental evaluation shows that XGBoost achieves a high prediction accuracy of 97.3%, significantly outperforming SVM, which attains 53.2% accuracy. The results confirm that the combined use of SVM and XGBoost provides a reliable and scalable solution for crime prediction.