Crime Data Analysis and Prediction Using Ensemble Learning

Abstract

Crime data analysis and prediction has become more popular and vital problem for the community. Identifying the patterns of crime actions of a place is very important to prevent crime. Law enforcement agencies can perform actively and respond earlier if they have better information about crime patterns.to come up this problem, this paper introduced N-ensemble learning technique for crime prediction. The experimentation were divided into three categories: statistical data visualization, base classification model and ensemble learning model. The first category, base classification model comprises Naive Bayes, J48 and Random Tree classification model. As experiment shows Random Tree achieved better accuracy of 82.0227%. Second category, ensemble learning model for combining base classification model. The ensemble learning were divided into two formats: the 1-ensemble model, the same type of classifier with diversified training sets and 3-ensemble model, combining three different base classifiers. The experimentation result showed that 1-ensemble model provides higher accuracy of 81.6073% and 3-ensemble model 79.2353%. Third category, statistical data visualization, analysis the impact of time, month and season on crime occurrence. Based on the experimentation more number of crime occurrence happened between the time intervals: 3:00pm to 6:59pm, and least number of crime occurrence between 3:00am to 6:59am compared to the other time slots. For month in July is the highest crime occurrences rate and least in February. The highest crime occurrence is observed during summer season compared to other seasons and least in winter. Crime occurrence decrease while moving from summer, autumn, spring and winter respectively.