<![CDATA[The growing use of communication media makes information easy to obtain including information on products provided in online stores. The rating feature on the website is a way to see the quality of the product to be purchased so as not to make a wrong choice when purchasing a product that can have a bad impact. Abundant data about product reviews in various online sources are useful as study material for producers in improving product quality. The existence of review data that is found without accompanying the rating makes it difficult for producers to determine the review into a particular sentiment. In this study can accelerate the determination of reviews into sentiment in the form of rating. This study uses a linear regression method and k-nearest neighbor as a prediction method and the method of weighting the contextual valence shifter based on the lexicon dictionary after pre-processing. The use of n-gram includes unigram, bigram, and trigram aimed at increasing the accuracy of the system. The greatest percentage is obtained at tolerance 1 with the results obtained by trigrams greater than bigram or unigram with linear regression method, namely 77% accuracy while k-NN gets 75% accuracy at k = 20 and k = 30. The test results show the use of n-grams, especially bigram and trigram has a positive impact on the results of system accuracy.]]>
Copyrights © 2019
