Proposal for a Model for Diabetes Detection Using Machine Learning Techniques

Carrera B.; Ccompi W.; Evangelista G.; Ticona W. · Lecture Notes in Networks and Systems (2024) · DOI: 10.1007/978-3-031-70518-2_14

Resumen

Currently a very relevant problem is diabetes, this disease is the cause of thousands of deaths and tends to grow in the coming years. The objective of the present research is to detect diabetes using Machine Learning techniques. For this purpose, the Indian PIMA database (PID) was used, which was extracted from Kaggle. Preprocessing techniques such as null value treatment and feature normalization were applied. Finally, various learning models such as Multilayer Perceptron, Neural Networks, Decision Tree, Support Vector Machines, K-Nearest Neighbors, Logistic Regression, Random Forest and Gradient Boosting algorithm were developed. From the best results obtained from the evaluation of the models, it was obtained that the best of them is the one generated by the Multilayer Perceptron Neural Networks and Support Vector Machine whose metrics were superior in Accuracy with 84.62% and Recall with 76.92%; however, in Precision it was surpassed by the Random Forest algorithm with 81.81%. It follows that both the Multilayer Perceptron and Support Vector Machine models accurately predicted the onset of diabetes, thus establishing their effectiveness in this predictive task, offering invaluable assistance to the healthcare sector. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.

Instituciones: Universidad ESAN; Universidad Tecnológica del Perú