Crime Forecasting Using Machine Learning Technique: The Case of Addis Ababa Police Commission

Abstract

As the population is increasing rapidly in the fast-growing world today, the number of crimes is also increasing from day to day. It takes a great deal of data analysis and model building to prevent future crimes while attempting to solve problems like crime using the evidence that is now available. Machine learning improves criminal analysis and helps to prevent and reduce crime. In the rapidly growing world of today, both the population and the quantity and diversity of crimes are rising constantly. The reason that prompted the researcher to work on crime estimation is that at this time, different crimes are committed in different cities, especially in Addis Ababa City, so it is necessary to forecast the reasons for these crimes. we will try to develop a long-term retention mechanism in the future Information should be produced from the data processing. Due to these issues, criminal evaluation and criminal analysis using machine learning will significantly reduce and prevent crime. Machine learning is one of the most often used in today’s technology. The purpose of this is study uses techniques to anticipate crime and create a crime forecasting model for the Addis Ababa Police Commission. For these cause Support vector machine (SVM), Neive Bayes, random forest classifiers, Decision tree, and AdaBoost classification models were utilized to achieve this goal. There was a total of 3539 datasets utilized in this study from the police commission. To test the performance of the prediction classification models, the researcher employs the confusion matrix, recall, precision, and roc-curve. After evaluation of the five classification models, the finding shows that all classification has the same accuracy of 97.879%. The researcher also proves that crime type, age, education level, Job, marital status, sex, and specific place were the shared population analysis factors of victims and offenders who were exposed to crime. The dare in this study was the anthropology similarity, of both victims and offenders. So, the study is to be completed for each alone.