A prediction model for criminal levels specialized in Brazilian cities

Abstract

The increase in violence around the world is becoming a major problem, causing severe damages to society: material, social and physical ones. The Government needs effective tools to fight against crime, and therefore, some tools are necessary to assist in the prevention of further crimes, in the allocation of its resources and visualization of geographic areas with high crime concentrations. This paper proposes a model of data mining, predicting criminal levels in urban geographic areas. The model was proposed to work using Brazilian data, specifically criminal and socio-economic ones. This work shows the approach proposed to face the problems of this social phenomenon, as a unified process to build a system which can able to help decision managers to fight and prevent crime. To validate the proposed procedure it was used as a case study. Using the crime and socioeconomic data of the Metropolitan Region of Fortaleza - Brazil (RMF). The case study proved that the process is useful and effective in building a predictor of criminal levels. The model achieves 70% of accuracy using an innovative method and heterogeneous data sets. © 2012 ICST Institute for Computer Science, Social Informatics and Telecommunications Engineering.