A Machine Learning-based Diagnosis and Prediction of Diabetes Mellitus Disease

Abstract

Traditional diagnostic techniques are prone to human error and time consuming. Computer-aided diagnostic techniques improve the performance and reduce the expenses. This paper presents machine learning based classifiers to detect diabetes in India and Indian Demographic & Health Survey (2019–21) dataset is considered for the analysis. Classifiers like Support Vector Machine, Decision Tree, Extreme Gradient Boosting, and Random Forest are considered. Principal Component Analysis is used for feature reduction. Hyperparameters are tuned to achieve good performance of the classifiers. The four phases adopted in the work include preprocessing, feature extraction, classification, and analysis. From the results, the Random Forest has given the maximum classification accuracy, precision, recall, and area under the curve in comparison with other models. The work finds application in healthcare for predictive analysis of diabetes.