In this article, we used the data from the system capture news site regarding safety, accidents, and other security occurrences. This system captures news security news from the region of Brusque. In this study, data were used from 01/03/2017 to 06/30/2017. Text mining techniques were used to get the incident address and Google API to get the latitude and longitude coordinates of the incidents. In this work, only the incidents classified as: Theft were used. In this database was applied a technique of analysis of data mining groupings known as: k-means in the geospatial coordinates, which has the purpose of grouping data more homogeneous among them and heterogeneous between the groups. With that, seven groupings or regions were obtained that show a map of the risk areas with seven clusters. Where it can be observed that the central region where the thefts are concentrated.