The Analysis of Attribution Reduction of K-Nearest Neighbor (KNN) Algorithm by Using Chi-Square

Abstract

Data reduction is one of the applicable techniques used to obtain the reduction representation from the data whose volume is much smaller, but still retains the original integrity of the data. Attribute reduction is a process to identify and eliminate an attribute with irrelevant or excessive values. In this study, attribute reduction was carried out using the Chi-Square Algorithm implemented in K-Nearest Neighbor (KNN) for classifying the objects based on the closest data to the objects.The test was carried out on the Pima Indians dataset with the total of 768 data. The Chi-Square method was used to reduce the dimensions of large datasets and to improve the accuracy of the prediction of the closest K-Nearest Neighbor results. K-Nearest Neighbor using Chi-Square as the choice of features proved to be accurate and effective in reducing data. The results of this study show that K-Nearest Neighbor using Chi-Square as the choice of features was accurate and effective in reducing data without reducing the integrity of the original data and reducing the quality of the information produced.