An Efficient Diabetic Prediction and Diagnosis Using Machine Learning Techniques

Author (s) : S. Ramesh Kumar and K. Aravind Kumar

---

Abstract :
Diabetes is the frequent and upward diseases in several countries. The patients undertake various preventive measures to avoid this disease during the beginning stage by predicting the symptoms. Diabetes can be predicted using several methods. The greatest challenge to current health care is to keep in control of the diabetes level of the patients. Diabetes is diagnosed using naïve Bayes algorithm. In this paper diabetic diseases prediction is implemented using apriori algorithm. A set of 16 parameters are used as the features for the prediction of diabetes. Apriori algorithm is used to reduce the features using the subsets. The popular Support Vector Machine (SVM), Naive Bayes (NB), Decision Tree (DT) algorithms are used for classification of diabetic and non diabetic and further the severity percentage of diabetes is predicted. Dataset used for the predictions is obtained from UCI repository. The overall Prediction accuracy for Diabetic Cases is 82.8% and for Non diabetic, the accuracy is 89.8%. The Diabetic Diagnosis overall accuracy for SVM is 37.8. Naive Bayes is 68.13%, Decision tree is 14.24%.

Keywords : Data Mining, Prediction, Diabetic, Apriori, SVM, Naive Bayes, Decision Tree.

Conference Date : 26-04-2019

Pages : 14-20

Paper ID : 4482455