Crime analysis and prediction using machine-learning approach in the case of Hossana Police Commission

Abstract

Crime is a socioeconomic problem that affects the quality of life and economic growth of a country, and it continues to increase. Crime prevention and prediction are systematic approaches used to locate and analyze historical data to identify trends that can be employed in identifying crimes and criminals. The objective of this study is to predict the type of crime that occurred in the city and identify the important features that make this prediction using a machine learning technique. For this experimental investigation, a supervised learning method was used to classify the types of crimes based on the final labelled class that indicates which type of crime is committed. Thus, decision tree (DT), random forest (RF), and K-nearest neighbor (KNN) algorithms are utilized along with the Python programming language in the Jupyter notebook environment. A total of 1400 records and nine attributes were used, and the data were split into training and testing sets, with 80% allocated to training and 20% for testing. The decision tree achieved an accuracy score of 84%, followed by the random forest at 86.07% and K-nearest neighbor at 81%. Besides this, the job of the offender, the victim’s age, and the offender’s age are the important features that cause crime. Therefore, it can be concluded that the use of machine learning to analyze historical data and the random forest algorithm to classify crimes yields promising results in predicting the type of crime.