Predictive Measurements of Diabetes Using Comparative Machine Learning algorithm

Diabetes is generally caused by the increasing quantity of glycemia. The world is now behind the different predictive algorithms to predict the different critically unsolved aspects. Predictive algorithms are generally based on machine learning or AI-based. These algorithms are used to solve different issues in healthcare also. In this article six different machine learning algorithms i.e. Logistic Regression, KNN (K Nearest Neighbor, Naïve Bayes (NB), Random Forest(RF), Decision Tree(DT), and Support Vector Machine(SVM) are used. Along with these algorithms, function elimination is done by via Boruta feature selection, treating the outliers with Quantile Transformer and extensive hyperparameter tuning using randomizedSearchCV and gridSearchCV. Finally, the result shows that SVM outperforms other algorithms with an accuracy of 93.116%, which is a feasible and realistic method for predicting diabetes. The PIMA diabetes dataset from the National Institute of Diabetes is used to investigate performance measurement. The overall performance is checked for all the algorithms, but the best performance is shown by SVM as 93.11%