Sentiment analysis on product review using support vector machine (SVM)

Public opinion can influence an organization or a company profile. It is important for the company to evaluate public response regarding their products. However, monitoring and organizing of public opinion are not easy. There are many published opinions in social media that is difficult to be processed manually. Therefore, a technique for automatically categorizing public reviews into positive or negative is needed. This research examined the classification of user reviews on products Windows Phone by implementing Support Vector Machine (SVM). This study used four different methods in tokenization stage: unigram, bigram, trigram, and n-gram. In the preprocessing data, there were 8 experiments whereas each group implemented 2 stemming algorithms (Snowball Stemmer and Iterated-Lovin Stemmer). The results were used as input of classification. Each input was classified 3 times with values C 0.25, 0.5, and 1.0. In conclusion, the result yielded the most appropriate model from n-gram with algorithm Iterated-Lovin Stemmer and C value 1.0. Finally, the use of SVM proves to be feasible approach of sentiment analysis regarding specific product.