Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{198800,
        author = {Mrs Y Latha and Mummoju Harshitha and Parne Srivani and Mohammed Mehraj Chowdhry and Koyalkar Priyanshu},
        title = {Crime Detection and Prevention System: A Full-Stack AI/ML Framework for Indian Crime Analytics},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {11},
        pages = {10985-10994},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=198800},
        abstract = {The increasing number of reported crimes and the insufficient effectiveness of standard police methods create ongoing obstacles for India to achieve its crime prediction and prevention goals. This research introduces a complete Crime Detection and Prevention System that uses machine learning and retrieval-augmented generation (RAG) technology and large language models (LLMs) to provide location-based crime analysis. The system operates on a dataset from the National Crime Records Bureau (NCRB) which contains 8,422 crime records and analyzes seven main crime categories after solving class imbalance issues. The Random Forest classifier achieves its highest accuracy of 83.95% to become the main predictive model which outperforms XGBoost. The RAG pipeline retrieves city- and state-specific crime knowledge to enhance contextual intelligence, which the Qwen 2.5-7B-Instruct LLM processes to produce structured predictions and reasoning and prevention strategies. The system combines NCRB statistical priors with rule-based logic to create a hybrid prediction system that maintains strength while losing functionality in different operational situations. The application operates as a full-stack solution which FastAPI uses for its backend and React for its frontend while allowing law enforcement officers and citizens to access specific features according to their roles. The system provides multiple functions which include real-time crime prediction and hotspot analysis and interactive visualizations and customized safety recommendations. The experiments showed strong results for different crime types, but they also revealed particular issues with categories that included both kidnapping and rape. The proposed system connects unprocessed crime information with operational intelligence, which assists organizations in making proactive security decisions.},
        keywords = {Crime Prediction, Machine Learning, Random Forest, XGBoost, Retrieval-Augmented Generation (RAG), Large Language Models (LLM), Qwen 2.5, NCRB Data, Predictive Analytics, Public Safety, FastAPI, React, Crime Analytics System},
        month = {April},
        }