2016 : A computational hybrid model with two level classification using SVM and neural network for predicting the diabetes disease

Prof. Ir. I Nyoman Sutantra M.Sc., Ph.D


Abstract

Data Mining is a collection of number of computational approaches. These approaches are used to develop Knowledge inference systems by identifying the hidden and convincing patterns from the input data. The aim of this study is to propose a computational Hybrid Prediction Model (HPM) for efficient diabetes prediction. The Pima diabetic dataset is used as the data source, obtained from the University of California, Irvine (UCI), the machine learning repository. At first stage of the proposed HPM the filtration feature selection method of MATLAB is used for selecting the most discriminatory predictors, reflecting the possibility of diabetes occurrence. At second stage, a two-layered classification is applied on the filtered data, by combining Support Vector machine (SVM) and Neural Network, to enhance the overall recognition rate of the model. The proposed hybrid model gained 96.09% of overall accuracy. The comparative study is also conducted and it is evident that the proposed model had obtained the significant classification accuracy. The accuracy rates achieved by many researchers in the past, on the same data set, ranges from 59.4% to 92% of accuracy. Further, for validating and evaluating the results, Recognition rate, Mean Absolute Error (MAE) and Receiver Operating Characteristics (ROC) performance measures are used. This research can be helpful to the physicians for predicting and detecting the diabetes at an early stage, efficiently.