Scholarlink

Crime prediction is essential for urban safety, utilizing machine learning (ML) and deep learning (DL) to analyze and forecast criminal activities. This study focuses on crime patterns in India using the Indian Crime Dataset (2020–2024) and introduces a hybrid modeling approach to enhance predictive accuracy. Our proposed model integrates deep learning techniques, such as long short-term memory (LSTM), with traditional time-series forecasting methods like ARIMA and Seasonal ARIMA (SARIMA). LSTM effectively captures temporal crime patterns, while SARIMA improves predictions by accounting for seasonal variations. We evaluate multiple ML models, including random forest, XGBoost, and ensemble techniques. Hybrid models demonstrate superior accuracy, achieving 97.6% (RF + SVM), 97.4% (RF + logistic regression), and 96.6% (XGBoost + KNN), outperforming standalone models. Exploratory analysis reveals crime hotspots in Delhi, Mumbai, Bangalore, and Hyderabad, with peak crime rates in the late morning and afternoon. This study highlights the potential of ML in crime prediction, offering valuable insights for law enforcement to optimize surveillance, deploy targeted interventions, and develop data-driven crime prevention strategies.

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

Crime Prediction and Forecasting in India Using Hybrid Machine Learning and Deep Learning Techniques

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Sudhanshu Vikram,
  • Arun Solanki

摘要

Crime prediction is essential for urban safety, utilizing machine learning (ML) and deep learning (DL) to analyze and forecast criminal activities. This study focuses on crime patterns in India using the Indian Crime Dataset (2020–2024) and introduces a hybrid modeling approach to enhance predictive accuracy. Our proposed model integrates deep learning techniques, such as long short-term memory (LSTM), with traditional time-series forecasting methods like ARIMA and Seasonal ARIMA (SARIMA). LSTM effectively captures temporal crime patterns, while SARIMA improves predictions by accounting for seasonal variations. We evaluate multiple ML models, including random forest, XGBoost, and ensemble techniques. Hybrid models demonstrate superior accuracy, achieving 97.6% (RF + SVM), 97.4% (RF + logistic regression), and 96.6% (XGBoost + KNN), outperforming standalone models. Exploratory analysis reveals crime hotspots in Delhi, Mumbai, Bangalore, and Hyderabad, with peak crime rates in the late morning and afternoon. This study highlights the potential of ML in crime prediction, offering valuable insights for law enforcement to optimize surveillance, deploy targeted interventions, and develop data-driven crime prevention strategies.