An Efficient Approach Parallel Support Vector Machine for Classification of Diabetes Dataset

paper proposes a Parallel SVM for predicting the diabetes chances in human based on a survey dataset which relates the different body parameters with diabetic and non diabetic persons. The aim of the paper is to correctly predict the future possibility of diabetes for any person. Since the survey dataset size could be very large with large numbers of parameters which makes it difficult to handle by simple SVM hence a parallel SVM concept is proposed in this paper to distribute these datasets into n different sets for n different machines which reduces the computational complexity, processing power and memory requirements for each machine. The proposed method is simple but quite reliable for parallel operation of SVM and can be used for large and unbalanced datasets the method also provide the flexibility to modify according to the dataset size, processors and memory available on different units. We have tested the proposed method using MATLAB and results are very encouraging.