Predictive Model for Diabetic Patients using Hybrid Twin Support Vector Machine

Twin Support Vector Machine (TSVM) is one of the most advanced machines learning approach in the area of pattern recognition. This paper proposes a predictive model based on Hybrid-TSVM (H-TSVM) for the classification of diabetic patients. The performance of the Twin Support Vector Machine is directly affected by the choice of kernel function. This research work combines two kernel functions, Polynomial and Gaussian kernel, to effectively utilize their strength. The resultant hybrid kernel function has better generalization and learning ability. The experiment has been performed on the PIMA Indian Diabetes data using 10-fold cross validation. The proposed predictive model is used to predict whether a new patient is suffering from diabetes or not. A comparative analysis of various predictive models based on PIMA Indian dataset has been performed and the accuracy of previously developed models has been compared with the proposed model and it is recognized that the performance is more accurate with the Hybrid TSVM.