Stratification for the Improvement of the Performance of an ANN in Diabetes Detection (#1308)

Authors

Patiño-Pérez, Darwin

Iñiguez-Muñoz, Felicísimo

Ochoa-Flores, Ángel

Córdova-Aragundi, José

Castro-Carrasco, José

Luque-Letechi, Alex

Munive-Mora, Celia

Abstract

One of the problems that have been detected in the generalization of machine learning models, which has been implemented for the detection of diabetes mellitus in the comprehensive private health diagnostic center of the city of Guayaquil; What prevents achieving optimal performance of the supervised learning model, for its implementation in the comprehensive center, is the imbalance or imbalance of the data depending on the type of class. For this reason, since the artificial neural network obtained the most acceptable performance, this study will focus on the evaluation of the artificial neural network and the improvement that has been achieved by applying the data stratification technique, which allows selecting from proportionally to the data groups in a balanced way, allowing to increase the performance of the artificial neural network by 3.39% but with a very high loss reduction of approximately 99.98% during learning.