This work presents the application of data mining techniques through predictive classification, using the application of data mining algorithms in an academic educational database of a higher education institution. Our objective was to identify patterns and classify students with the most prone to evasion profile and to discover the possible reasons that contribute to the growth of evasion. The results of the experiments show that: failing in the four basic subjects of the course, not participating in any type of project, together with the extrapolation of the 8 normal semesters of the course and having an age group over 26 years, are the factors that collaborated most for the course dropout.