Towards to a Predictive Model of Academic Performance Using Data Mining in the UTN - FRRe

David L. La Red Martínez, Marcelo Karanik, Mirtha Giovannini, Reinaldo Scappini

Students completing the courses required to become an Engineer in Information Systems in the Resistencia Regional Fac-ulty, National Technological University, Argentine (UTN-FRRe), face the chal-lenge of attending classes and fulfilling course regularization requirements, often for correlative courses. Such is the case of freshmen's course Algorithms and Data Structures: it must be regularized in order to be able to attend several second and third year courses. Based on the results of the project entitled “Profiling of students and academic performance through the use of data mining”, 25/L059 - UTI1719, implemented in the aforementioned course (in 2013-2015), a new project has started, aimed to take the descriptive analysis (what happened) as a starting point, and use advanced analytics, trying to explain the why, the what will happen, and how we can address it. Different data mining tools will be used for the study: clustering, neural networks, Bayesian networks, decision trees, regression and time series, etc. These tools allow differ-ent results to be obtained from different perspectives, for the given problem. In this way, potential problematic situations will be detected at the beginning of courses, and necessary measures can be taken to solve them. Thereby, the aim of this projects is to identify students who are at risk of abandoning the race to give special support and avoid that situation. Decision trees as predictive classification technique is mainly used.

Full Text