The concentration of police resources in conflict zones contributes to the reduction of crime in the region and the optimization of those resources. This paper presents the use of regression techniques to predict the number of criminal acts in Colombian municipalities. To this end, a set of data was generated merging the data from the Guardia Civil with public data on the demographic structure and voting trends in the municipalities. The best regressor obtained (Random Forests) achieves a RRSE (Root Relative Squared Error) of 40.12% and opens the way to keep incorporating public data of another type with greater predictive power. In addition, M5Rules were used to interpret the results.

No abstract

During the transit of students in the acquisition of competencies that allow them a good future development of their profession, they face the constant challenge of overcoming academic subjects. According to the learning theory, the probability of success of his studies is a multifactorial problem, with learning-teaching interaction being a transcendental element (Muñoz-Repiso and Gómez-Pablos in Edutec. Revista Electrónica de Tecnología Educativa 52: a291-a291 (2015), [1]. This research describes a predicative model of academic performance using neural network techniques on a real data set.

No abstract

The analysis of criminal information is critical for the purpose of preventing the occurrence of offenses, so the crime records committed in the past are analyzed including perpetrators. The main objective was to identify crime patterns in the city of Bogota, Colombia, supported using Early Warning System based on data mining (CRISP-DM method). The research results show the identification of 12 different criminal profiles demonstrating that the Early Warning System is applicable since it managed to significantly reduce the time devoted to the processes of registering complaints and searching for criminal profiles.