To effectively support law enforcement agencies and improve public safety, it is essential to understand
patterns and trends in criminal activity. In this project, crime data collected from publicly available
sources and structured datasets is analyzed to identify patterns related to location, time, and
environmental factors. The primary objective of the system is to predict the category of crime that is
most likely to occur based on input features such as district, area type, weather conditions, crowd level,
lighting condition, police presence, and time-related attributes. Machine learning techniques are used
to build predictive models capable of analyzing historical crime data and generating accurate crime
predictions. Multiple classification algorithms, including Random Forest, Gradient Boosting, XGBoost,
and Support Vector Machine, are evaluated to determine the most effective model for prediction. The
system is implemented as a web-based application using the Flask framework, allowing users to input
relevant parameters and obtain predicted crime categories along with confidence scores and associated
risk levels. In addition, the system applies clustering techniques such as K-Means to identify crime
hotspot zones based on geographical coordinates. By analyzing spatial and temporal crime patterns,
the system provides meaningful insights that can assist authorities in planning preventive measures and
improving crime monitoring strategies. Finally, the results obtained from different machine learning
models are compared and analyzed, and possible directions for further improvement of the system are
discussed.
