An Effective Disease Prediction System using CRF based Butterfly Optimization, Fuzzy Decision Tree and DBN

Abstract

Diabetes is a seriously deadly disease today. It is necessary to enable patients to control their blood glucose levels. Even though, in the past, various researchers proposed numerous diabetic detection and prediction systems they are not fulfilling the requirements in terms of detection and prediction accuracy. Nowadays, diabetes patients are utilizing the gadgets like Wireless Insulin Pump that passes into the body instead of syringes for filling insulin. Within this context, insulin treatment is necessary for avoiding life-threatening. Toward this mission, a new deep learning approach-based disease detection system is introduced which takes care of identifying Type-1 and Type-2 diabetes, heart diseases, and breast cancer. In this system, a new Conditional Random Field based Butterfly Optimization Algorithm (CRF-BOA) is developedto select the important features for identifying the Type-1 and Type-2 diabetic disease. Besides, a new fuzzy ID3 classification method is developed for classifying the patient's datasets either normal or abnormal and disease affected. Ultimately, by applying the deep belief network (DBN) the classified patient records are involved with training to identify the relevant symptoms of similarity and glucose status of various patient records. These experiments are being conducted for proving the efficiency of the proposed deep learning approach in terms of glucose monitoring efficiency and disease prediction accuracy.The proposed approach achieved high detection accuracy than the current deep learning approaches in this directionbased on error rate and accuracy.