CDAPS – Crime Data Analysis & Prediction System Using Time Series Forecasting

The growing availability of crime-related data has opened up significant opportunities for creating smart analytical systems that support proactive law enforcement strategies. However, many traditional crime management systems still rely mostly on historical reports and descriptive statistics, which provide limited help in predicting future crime trends or spotting new criminal activities. To tackle these issues, this research proposed a Crime Analytics and Forecasting Framework that combines machine learning, time-series forecasting, anomaly detection, and interactive visualization into a single decision-support platform.  This framework was designed to turn raw historical crime records into useful predictive insights through a systematic process that includes data collection, preprocessing, feature engineering, model training, forecasting, anomaly detection, and visualization. Historical crime data collected from various Indian cities were cleaned, combined, and turned into a structured dataset for predictive analysis. Different feature engineering techniques, such as temporal feature extraction, cyclical encoding, lag variables, and rolling statistical measures, were used to capture seasonal behavior and historical dependencies in the dataset. These preprocessing steps significantly improved the dataset's quality and boosted the learning ability of the prediction models used.  To assess the effectiveness of different predictive methods, four ensemble machine learning algorithms—Random Forest, Gradient Boosting, XGBoost, and LightGBM—were implemented and compared for crime estimation tasks. Additionally, three forecasting techniques, SARIMA, Prophet, and Long Short-Term Memory (LSTM), were tested to find the best model for predicting future crime incidents. The comparative analysis showed that the SARIMA model consistently delivered reliable forecasting results by effectively capturing seasonal trends and repeating crime patterns found in the historical data. Among the machine learning models, Gradient Boosting performed best due to its ability to iteratively reduce prediction errors and model complex nonlinear relationships within structured crime data.  Along with forecasting, using statistical anomaly detection methods further improved the framework's capabilities. Implementing the Z-Score and Interquartile Range (IQR) methods allowed the system to identify unusual fluctuations and unexpected spikes in crime activity that may require immediate action from law enforcement agencies. Recognizing such anomalies can help authorities identify strange events, investigate possible security threats, and take timely preventive measures. The visualization module, built with a FastAPI-based backend and an interactive dashboard, successfully converted complex analytical outputs into easy-to-understand graphs, charts, and comparative reports, making the forecasting results available to both technical users and decision-makers.  In summary, the proposed framework shows that combining machine learning, statistical forecasting, anomaly detection, and visual analytics can significantly enhance crime intelligence and support data-driven decision-making. Instead of just being a crime prediction model, the framework acts as a complete analytical platform that can help law enforcement agencies, policymakers, researchers, and smart city administrators understand historical crime behavior, anticipate future trends, and optimize resource allocation. The system’s modular design also ensures scalability, allowing for the addition of more analytical components and data sources with minimal changes. Therefore, this framework represents a practical step toward building intelligent, data-driven public safety systems that can support proactive policing and urban crime management.

[1] M.-S. Baek, W. Park, J. Park, K.-H. Jang, and Y.-T. Lee, "Smart Policing Technique With Crime Type and Risk Score Prediction Based on Machine Learning for Early Awareness of Risk Situation," IEEE Access, vol. 9, pp. 131906–131921, 2021. [2] S. Hossain, A. Abtahee, I. Kashem, M. M. Hoque, and I. H. Sarker, "Crime Prediction Using Spatio-Temporal Data," in Communications in Computer and Information Science (CCIS), Springer, vol. 1235, pp. 277–289, 2020. [3] S. Sridharan, S. N., V. S., and P. Santhi, "Crime Prediction Using Machine Learning," EAI Endorsed Transactions on Internet of Things, vol. 10, 2024. [4] G. García-Zanabria, M. Angelini, N. Ferreira, and C. Turkay, "CriPAV: Street-Level Crime Patterns Analysis and Visualization," IEEE Transactions on Visualization and Computer Graphics, vol. 28, no. 12, pp. 4000–4016, 2022. [5] S. Tuarob, S. Chotiprapa, and M. H. Ashaduzzaman, "Beyond Administrative Reports: A Deep Learning Framework for Classifying and Monitoring Crime and Accidents Leveraging Large-Scale Online News," Neural Computing and Applications, vol. 37, pp. 7183–7205, 2025. [6] R. J. Hyndman and G. Athanasopoulos, Forecasting: Principles and Practice, 3rd ed. Melbourne, Australia: OTexts, 2021. [7] G. E. P. Box, G. M. Jenkins, G. C. Reinsel, and G. M. Ljung, Time Series Analysis: Forecasting and Control, 5th ed. Hoboken, NJ, USA: Wiley, 2015. [8] S. Makridakis, E. Spiliotis, and V. Assimakopoulos, "Statistical and Machine Learning Forecasting Methods: Concerns and Ways Forward," PLOS ONE, vol. 13, no. 3, 2018. [9] L. Breiman, "Random Forests," Machine Learning, vol. 45, no. 1, pp. 5–32, 2001. [10] J. H. Friedman, "Greedy Function Approximation: A Gradient Boosting Machine," The Annals of Statistics, vol. 29, no. 5, pp. 1189–1232, 2001. [11] T. Chen and C. Guestrin, "XGBoost: A Scalable Tree Boosting System," in Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016, pp. 785–794. [12] G. Ke, Q. Meng, T. Finley, T. Wang, W. Chen, W. Ma, Q. Ye, and T.-Y. Liu, "LightGBM: A Highly Efficient Gradient Boosting Decision Tree," in Advances in Neural Information Processing Systems (NeurIPS), 2017.