<![CDATA[Digital distribution service is a container for various applications that can be downloaded at any time. In addition to applications on Digital distribution service, there are also application reviews that contain comments from certain application users. The review contains a very large number of negative comments or positive comments. Due to the large number of reviews, the digital distribution service shares these reviews using ratings with inappropriate review content. To solve the problem of mismatch between the content of the review and the rating given by the user, a sentiment analysis is needed. This study uses the Naive Bayes method and the Modified Term Frequency Scheme. Naive Bayes method was chosen because it works well in document classification by estimating the required parameters. Used 1,500 data consisting of 627 positive reviews and 873 negative reviews. Preprocessing process is carried out, weighting using the Modified Term Frequency Scheme and document classification using the Naive Bayes method. In the 5-fold test, the average of the method used was accuracy 83%, recall 86%, precision 76%, f-measure 77,70% with the 3rd fold being the best fold with accuracy 85%, recall 84,50%, precision 81,34%, f-measure 82,88%.]]>
Copyrights © 2021
