A stratified sampling technique based on correlation feature selection method for heart disease risk prediction system

In medical, data mining method can be utilized by the physicians to improve clinical diagnosis. In this paper a stratified approach named Correlation Feature Selection Stratified Sampling (CFS-SS) has been introduced. This method is applied to medical diagnosis heart disease risk prediction system. By using this proposed system the attributes are grouped together into homogenous sub groups, before sampling the strata will be mutually exclusive, every attribute will be assigned to only one stratum. The original dataset is given to the filter Correlation based feature selection (CFS) system. The output of the system will be the efficiently achieved without stratified sampling. Stratified sampling of all the sub sets are put together in such a manner that subset of same group size will be in one group by using CFS-SS. The efficiency of proposed system (CFS-SS) is better than existing system (CFS). Key wordsCFS, CFS-SS, SVM, NB & DT.