Implementasi Text Mining Pada Analisis Sentimen Pengguna Twitter Terhadap Marketplace di Indonesia Menggunakan Algoritma Support Vector Machine

In this digital era, technology development has changed the behavior of society from buy offline to online. One of this behavioral changes is marked by the growth of global marketplace including in Indonesia. The big marketplaces in Indonesia that have received a lot of public response on social media are Tokopedia, Shopee, and Bukalapak. This research determines the public sentiment toward both the service and issues surrounding these three marketplaces on media social especially Twitter. Public opinion is classified into a positive or negative sentiment. The data used in this study is obtained from Twitter API (Application Programming Interface) using keyword Shopee, Tokopedia, and Bukalapak. Preprocessing texts are divided into five steps: cleansing, case folding, stemming, stopwords, and tokenizing. Training and testing data are divided using k -fold cross validation method, while visualization the characteristic of text is using word cloud. Research shows that public are posting tweet more positive sentiment than negative one. The perfomance of classification shows that the best G -mean and AUC value for Bukalapak testing data are 0.85 and 0.86 in the first fold. While the best G -mean and AUC value for Shopee testing data are 0.76 and 0.77 in the seventh fold and the best G -mean and AUC value for Tokopedia testing data are 0.82 and 0.83 in the sixth fold. Keywords : sentiment analysis, marketplace, support vector machine, twitter