A Comparative Analysis of Diabetes Risk Prediction Techniques Using Data Analytics

Diabetes is one life-threatening disease. Instead of the physical examination or medical test of diabetes, Initial screening based on the characteristics is a more critical element. Automatic detection of diabetes prediction helps for accessing the health of the person. In this study, Machine Learning (ML) techniques are used to predict the risk of diabetes. ML methods such as Naïve Bayes, K-NN (Nearest Neighbour), Support Vector Machine (SVM), Random Forest (RF) are performed. Two secondary datasets are used to evaluate the performance of the ML Techniques. The wrapper stepwise Feature Selection (FS) method is applied to find the critical attribute for classifying diabetes patients. Finally, Machine Learning Score is arrived with the combination of accuracy and number of features extracted. The outcome of this analysis indicates that the SVM and K-NN obtains higher accuracy with minimum number of features.