The effect of feature reduction techniques on diagnosis of diabetes

Abstract

In this study, Bayesian and, Decision Trees classifiers are used for the automatic diagnosis of the diabetes disease. 17 attributes of the diabetics has been reduced to 4 attributes using principal component analysis and sequential forward selection algorithm. The performances of the classifiers obtained from the use of the dimension reduction techniques are compared. Thus, dimension reduction methods to examine the positive effects on both the results and is intended to reduce the workload of the machine learning. End of the study, it has been seen that Decision Trees Algorithm provides the highest performance criterion and Principle Component Analysis gives the best classifying results. The study produces the importance of the dimension reduction techniques to process the big demensional datas.