Multilayer Neural Networks for Predicting Academic Dropout at the National University of Santa - Peru

Investigations have applied the Machine Learning to predict whether a college student culminate or not his studies, however, in each scenario the factors that influence student graduation are multiples, then: how predict defection students at the U niversidad N acional of Santa - Peru with a precision greater than 90%? In the present research a model based on Multilayer Neural Networks was trained to predict the academic dropout at the School of Engineering from the aforementioned university, to Neural Networks Multilayer of 6 layers, it provided a model with an accuracy of the 98.97% in the training set, which is satisfactory in relation to alternative models they worked in 15 different experiments and which were compared with classification algorithms obtained in the service AutoAI of IBM Watson Studio that returned to classifier XGB as the best predictor with an accuracy of 87.1 %.