This article proposes the modified KNN (K Nearest Neighbor) algorithm which considers the feature similarity and is applied to the index optimization. The texts which are given as features for encoding words into numerical vectors are semantic related entities, rather than independent ones, and the index optimization is able to be viewed into a classification task where each word is classified into expansion, inclusion, and removal. In the proposed system, each word in the given text is classified into one of the three categories by the proposed KNN algorithm, associates words are added to ones which are classified into expansion, and ones which areclassified into inclusion are kept by themselves without adding any word. The proposed KNN version is empirically validated as the better approach in deciding the importance level of words in news articles and opinions. The significance of this research is to improve the classification performance by utilizing the feature similarities.