Data mining tools for prospective students’ success rate prediction

The aim of that study was prospective students’ success rate prediction in case if they are enrolled at university. For this purpose different classification and prediction methods were used namely support vector machines (SVM), neural networks and fuzzy systems which were obtained by using genetic and bionic algorithms and their modifications. Firstly, meta-heuristic called Co-Operation of Biology Related Algorithms (COBRA), that has earlier demonstrated its usefulness on CEC’2013 real-valued optimization competition benchmark and its modification for solving constrained optimization problems, was applied to SVM-based classifiers design. Then genetic algorithm was used for neural networks and fuzzy logic systems automatic generation. Various benchmark classification problems were solved with those approaches. It was established that support vector machines, neural networks and fuzzy logic systems developed in that way outperform many alternative methods on mentioned benchmark classification problems. So the workability and usefulness of proposed classification or prediction algorithms were confirmed. After that for solving of prospective students’ success rate prediction problem, in case if they are enrolled at university, information about them obtained during the operational period of Admission Committee was gathered and preprocessed. Eventually it was established which kind of information about a prospective student is enough for determining whether he or she will or will not pass end of first semester exams in case of enrollment. Moreover, it should be noted that support vector machines generated by collective optimization method COBRA have shown the best results.