Comparative Study of Multivariate Statistical Methods for Predicting the Academic Performance of Students at the University of Guayaquil

Universities are responsible for the students who have completed their career to face the labour field, their training or skills development is based on the knowledge acquired in them. The study of academic performance is the most importance and at the same time quite complex. It does not know what factors were causing low performance, predict what they were, and address them with better decisions. The study factors identified are academic and demographic, the latter is gender, age, the place where they live, and there are students who do not have the adequate infrastructure for a better development of their academic activities. Applying multivariate statistical techniques comparisons discover a better prediction model of academic performance in students of the University of Guayaquil. The main objective is to find patterns of academic performance in students, grouped by subject and semester, using comparisons of multivariate techniques. Applying classification models such as K-nearest-neighbors with an accuracy of 97.7%, best selection in decision trees with 99.54% and logistic regression with a probability belonging to a class of 97%. For the selection of the best technique, the best prediction was compared and the best prediction was selected.