Main Article Content
Diabetes is one of the diseases in the world that is spreading like epidemics. Every generation, ranging from girls, teenagers, young people and the elderly, is seen to suffer from this. In terms of organ failure such as liver, kidney, heart, stomach, prolonged impact can cause worse effects and can lead to death. It is commonly associated with retinopathy and neuropathy disorders. There are primarily two types of diabetes: type 1 and type 2.Diabetes diagnosis or prediction is carried out by various techniques of data mining, such as correlation, grouping, clustering and pre-processing of data. The research led to similar open problems to recognize the need for a relationship between the key factors contributing to the development of diabetes. This is possible through the mining patterns contained in the dataset between the independent and the dependent variable. This paper compares the accuracy of pre-processing of different models of dimensionality reduction. The process of DR is classified into two stages in the proposed work, one is unsupervised DR and other is supervised DR. before the processing using unsupervised DR improved principle component analysis has been done. Then the two dataset has been merged. On whole the aim for the proposed technique is to reduce the dimensionality which could improve the accuracy in feature selection. The simulation results gives the improved accuracy in pre-processing part.