Mobile application based on KDD to predict high-crime areas and promote sustainability in citizen security in a district of Lima-Perú

Metropolitan Lima faces a serious citizen security situation, reflected in high rates of crime and violence in several districts. The development of a mobile application to identify and predict areas of high crime incidence is proposed. Using historical data of criminal incidents and reports registered by users in the application, models capable of predicting the occurrence of crimes in real time are trained. The data mining process follows the KDD methodology, which includes the stages of selection, preprocessing, transformation, data mining, evaluation and knowledge consolidation. Machine learning algorithms, such as Random Forest and Gradient Boosting, were used to make these predictions. Visualization techniques, such as heat maps, were also used to represent crime events and facilitate their understanding by users. The results show an accuracy of 88% for the Random Forest algorithm and 91% for Gradient Boosting in predicting the occurrence of crimes, which demonstrates the effectiveness of machine learning models to improve citizen security in Metropolitan Lima, therefore these findings have significant implications for crime prevention and suggest that the application of these technologies can be fundamental to address security challenges in the city.