<![CDATA[In online stores there are reviews of items that contain comments about feedback from previous buyers that are useful for subsequent buyers as well as sellers at online stores. Reviews Usually consist of negative comments or positive comments. The number of reviews is very much. In overcoming this problem, sentiment analysis is needed. This study uses the Learning Quantization Vector and Term Frequency-Modified Inverse Document Frequency methods. The LVQ method was chosen because it has the advantage of being able to summarize the dataset into a codebook vector. The data used consisted of 250 positive comments and 250 negative comments. The data will be preprocessing, weighting the word using TF-mIDF and consequently using the LVQ method. The results of testing the LVQ parameters obtained an accuracy value of 75.11%, recall of 75.11% precision of 77,80%, f-measure of 76.43% with parameter values ​​of learning rate 10-3, dec α 10-6, and values maximum epoch 19. Based on the final test results, obtained the value of the Learning Vector Quantization method with TF-mIDF resulted in an average accuracy of 72.47%, recall of 72.47%, precision of 76.39%, and f-measure of 74.33 % and using the Learning Vector Quantization method with TF-IDF resulted in an average accuracy of 54.80%, recall of 54.80%, precision of 54.30%, and f-measure of 52.61%.]]>
Copyrights © 2019
