Evaluate the Predictive Performance of Supervised Machine Learning Algorithms in Diabetes Dataset

The progression of diabetes disorder is very alarming in our society nowadays.Not only increase the glucose level of a human body but also spread other complexities like blood pressure, heart disease, kidney disease and many more diseases that have the relation with diabetes.Still now most of the people are not aware of doing diagnosis and sometimes it is hard to find early also. Diagnosis of diabetes is more important to prevent or control this disease, so that machine learning might play a very significant role in this area because of its less error and efficiency. In order to predict diabetes, the author of this research attempted to evaluate the effectiveness of several machine learning (ML) methods, including Random Forest, SVM, Decision Tree, K-Nearest Neighbors, and Logistic Regression with the help of Pima Indians Diabetes Dataset from Kaggle repository. For each model confusion matrix has been used to measure Accuracy, Precision, Recall, and F1-Score for evaluating the performance. Random Forest got the highest accuracy of 81% while Decision Tree got the least accuracy of 69%. The methodology proposed in this paper is very easy to understand for everyone which is one of the major key points of this research.