Diabetes Detection and Forecasting using Machine Learning

Abstract

Diabetes Mellitus is a major health concern globally, requiring early diagnosis for its proper treatment. This research seeks to analyze the effectiveness of various machine learning algorithms for detecting diabetes using datasets from the Behavioral Risk Factor Surveillance System (BRFSS). In this context, two datasets have been considered for analysis: the binarized unbalanced dataset (BRFSS2015) and its balanced counterpart. The machine learning algorithms considered in this study are Logistic Regression, Random Forest, Gradient Boosting, K-Nearest Neighbors, Decision Trees, XGBoost, Support Vector Machines (SVM), and Deep Neural Networks. Therefore, the Gradient Boosting classifier has achieved an accuracy of 96.75% on unbalanced data and 85.36% on balanced data, the best among the machine learning algorithms. However, the Deep Neural Network algorithm has achieved 96.86% accuracy on the unbalanced dataset and 85.16% on the balanced dataset.