Predicting Life Style of Early Diabetes Mellitus using Machine Learning Technique

Abstract

A branch of artificial intelligence called Machine Learning (ML) enables machines to learn without having to be emphatically instructed. Machine Learning Techniques (MLT) have been used to forecast a variety of chronic diseases in the healthcare sector. Improvement in clinical approaches is necessary for early diabetes prediction to prevent complications and prolong the diagnosis of diabetes. Diabetes is growing fast in this world. In this paper MLT based Framework is recommended for early prediction of Diabetes Mellitus (DM). In this Paper the authors make use of PIDD data set. Different MLTs are used including Support Vector Classification (SVC), Logistic Regression (LR), K Nearest Neighbor (KNN) and Random Forest (RF). Data analysis is the first step in our method after which the information is transferred for data pre-processing and feature selection methods. RF performed better than other models with a 92.85 % accuracy rate followed by SVC (91.5%), LR (83.11) and KNN (89.6). K-fold cross-validation technique is utilized to verify the outcomes. The contribution of lifestyle characteristics is calculated using a feature engineering process. As a result, comprehensive overall comparative assessments of all the algorithms are performed taking into account variables such as accuracy, precision, sensitivity, recall, F1 score and ROC-AUC. The medical field can use the proposed framework to make early diabetes predictions. Additionally, it can be applied to other datasets that have data in common with diabetes.