<![CDATA[K-Nearest Neighbors (K-NN) is one classification method that easy to learn. Although, this method has some drawbacks, one of them is this classification could provide a low accuracy casued by a large feature space with irrelevant features among them. Because of that drawback, feature selection is applied to reduce the feature space by reducing number of irrelevant features in text classification. Selection feature method that being used in this experiment is using Chi-Square method. Using Chi-Square method to select important features by measuring dependency level of each feature across classes and documents. The process including in this experiment is collecting training and testing documents, text preprocessing and feature selection, and classification. After classification is being done by the system, we make an observation and analysis towards classification result, including precision, recall, and F-Measure value. From 16 evaluations, the best precision and recall score obtained with 90% precision and 78% recall on k = 15 using 25% feature selection used. While the best F-Measure score obtained with 78% F-Measure on k = 15 and k = 5 using 25% feature selection used. From this experiment, its appear that feature selection take effect in increasing F-Measure value in text classification of SAMBAT Online complaint documents in bahasa using K-Nearest Neighbors classification method.]]>
Copyrights © 2017
