Prediction of Crime Hotspots using Machine Learning with Stacked Generalized Approach

Crime prediction has an important part in growth of police tactics and the application of instruments for crime prevention and control. These days, machine learning is the most widely used prediction approach. Few research, nevertheless, have systematically examined several machine learning techniques for crime prediction. This study investigates the prediction capacity of various machine learning approaches employ historical data on public property crime from a portion of a huge seaside metropolis in India. Based just on historical crime data, the ADABOOST model outperforms K-nearest neighbour (KNN), Random Forest, Support Vector Machine (SVM), Naive Bayes, and Convolutional Neural Networks (CNNs). Information about the offenses and their consequences. Furthermore, the density of the metropolitan road network and the built-environment data of sites of interest (POIs) were included as covariates for the Boosting model. When historical crime data was the only basis for the prior model, the model that included covariates linked to the built environment had a significant predictive impact. Therefore, it is appropriate to employ both factors connected to criminological theories and historical crime data is used to forecast future crime. When it comes to predicting crimes, machine learning techniques vary from one another.