CRIME RATE PREDICTION AND ANALYSIS USING LSTM ALGORITHM

Routine Activity Theory explains that crimes happen when three elements come together: a motivated offender, a suitable target, and no capable guardian. Rational Choice Theory adds that offenders weigh risks, effort, and rewards before acting. Crime Pattern Theory combines these ideas, showing that offenders rely on familiar areas (their "cognitive maps") when choosing crime locations. Crime hotspots emerge in places that either create or attract criminal opportunities, influenced by past crimes and environmental factors. The proposed crime prediction system combines machine learning algorithms like Random Forest, KNN, SVM, and LSTM. Initially, only historical crime data are used to train these models and identify the best-performing one. Later, environmental factors such as road density and points of interest (POI) are added to assess if prediction accuracy improves. KNN classifies data by finding the nearest neighbors based on input features, using majority or weighted voting. Meanwhile, Naive Bayes (NB) predicts outcomes using probabilities and assumes that input features are independent, making it a simple yet effective classification method.