Performance of Supervised Algorithms in Predicting Academic Performance in University Students

Peralta Ascue M.; Merma Aroni J.L.; Santana Carbajal O.I.; Osco Mamani E.F.; Ale Nieto T.F. · CICIC 2023 - Decima Tercera Conferencia Iberoamericana de Complejidad, Informatica y Cibernetica en el contexto de the 14th International Multi-Conference on Complexity, Informatics, and Cybernetics, IMCIC 2023 - Memorias (2023) · DOI: 10.54808/cicic2023.01.79

Resumen

The purpose of this study is to evaluate the performance of supervised algorithms based on data mining techniques: Decision Trees and Neural Networks, in predicting the academic performance of students in the Business Intelligence course at the School of Systems Engineering and Computer Science at the Technological University of the Andes, Abancay-Peru. The CRISP-DM methodology was used for the data mining process modeling, and the Weka software was used as a tool for training and testing the algorithms. The supervised prediction algorithms studied and analyzed include J-48 Decision Trees, Random Forest, DEPTree, and Multilayer Perceptron. The performance of the algorithms was evaluated using accuracy, precision, and sensitivity indicators. It was observed that the J-48 algorithm had the highest performance in classification and prediction, with a precision rate of 91.67%, accuracy of 64.52%, and sensitivity of 52.38%, outperforming the other studied algorithms.

Instituciones: Universidad Nacional Micaela Bastidas de Apurímac; Universidad Privada de Tacna; Universidad Nacional Jorge Basadre Grohmann; Universidad Tecnológica de los Andes