Heinous Crime Prevention and Prediction Using Data Mining Techniques

Crime is dominant and unpredictable, may happen any place at any time, and is thus a difficult problem for any society to resolve, and predicting the crime before happening is a complex task. The present analysis of comparison for seven well-known prediction algorithms; Logistic Regression, Support Vector Machine, Decision Tree, K-Nearest Neighbor, Naive Bayes, Random Forest, and Stochastic Gradient Descent has led to the suggestion for a better crime prediction model. Using a crime dataset, Exploratory Data Analysis (EDA) has been carried out to find patterns and comprehend trends in crimes. Among the aforementioned algorithms, Logistic Regression outperforms with 0.85% accuracy. However, with an ensemble of the Logistic Regression, Decision Tree, and Support Vector Machine, the model has gained the accuracy of 0.86%. Our system has identified regions with a great possibility of crime incidence and can envisage crime-prone zones through effective patrolling and transfer posting of police officials. This prediction model assists the security agencies in using resources effectively, foreseeing crime at a certain time, day, month, year, and category, which provides expected policing by the department.