Predição de localização de crimes em região urbana usando algoritmos de análise de regressão

Abstract

There are relevant rates of violence in Brazil that have increased in recent years. Intelligence and efficiency are required to combat this issue, in order to reduce public money spending and time from public oficials. There are several solutions, this work also presents one of them, which use intelligent systems to predict where and when a crime will occur, this allows organizing police routes to areas with a higher risk of danger. In the experiments carried out, two databases were used, one from the city of Philadelphia, publicly available, and another from the city of Fortaleza, provided by the Department of Public Safety. For these experiments, different regression methods were applied to make predictions of the places where crimes could occur. For the dataset of the city of Philadelphia, only one experiment was performed using these regressors. For the dataset of the city of Fortaleza, four sets of experiments were carried out using the same regression methods used previously for the city of Philadelphia. For each of these techniques, a result was generated regarding the residues, which is the difference between the predicted values and the actual values. The use of regression methods also provided the generation of point dispersion plots, where each prediction made is plotted and compared to the original points. There was also a need to display the predicted points on the maps of the respective cities, so, this way, it is possible to check the areas where there are major criminal incidents. For each of the regressor methods, the error value was calculated using the metrics MSE (Mean Squared Error) and RMSE (Root Mean Squared Error), the lowest values for MSE and RMSE allow to infer that the model presented excellent predictions. The results obtained by the regression methods proved to be effcient in the task of predicting the location of crimes. It is possible to conclude that the use of these methods for problems of criminal aspects, make the predictive task much more tangible.