<![CDATA[This study aimed to analyze the effect of Mutual Information feature selection on classification performance in a boarding house decision support system. The study used the MamiKost dataset from Kaggle, which contains price, rating, and boarding house facility attributes. Price and rating were used to construct recommendation labels, while facility attributes were used as classification features. The research stages included data preprocessing, Mutual Information score calculation, feature subset selection, and classification performance evaluation using accuracy. The results showed that air conditioning achieved the highest Mutual Information score (0.263), followed by sitting toilet (0.147) and bathroom (0.026). These three features were selected as the feature subset. However, feature selection did not improve classification performance. Accuracy decreased from 0.9045 to 0.8909 after feature selection was applied. The findings indicate that features with low Mutual Information scores still contributed useful information to the classification process. Therefore, using all available features produced better performance than using the selected feature subset on the MamiKost dataset.]]>
Copyrights © 2026
