CRISP-DM-Based Mobile Application for Predicting High-Crime Areas in Metropolitan Lima

Abstract

The city of Lima, Peru, has been facing a serious climate of citizen security that has risen extremely high in recent years. The objective of this work is to identify and predict areas of high crime incidence through a mobile application based on historical data on criminal incidents recorded by users. The mobile application has been implemented using the CRISP-DM methodology, which includes the stages of business understanding, data understanding, data preparation, modeling, evaluation, and implementation. The main machine learning algorithms used were Random Forest and Gradient Boosting; likewise, visualization techniques such as heat maps were used to represent criminal events. The results obtained in the prediction of the occurrence of crimes were: Using the Random Forest algorithm, an accuracy of 87% was achieved and using Gradient Boosting 84%, These findings allow people who use the mobile application to know in real time which zones or areas are of high crime incidence therefore dangerous in this way they will be able to opt for prevention behaviors and that these technologies can help address the Security challenges in the city of Lima.