KLASIFIKASI DATA TWITTER PELANGGAN BERDASARKAN KATEGORI MYTELKOMSEL MENGGUNAKAN METODE SUPPORT VECTOR MACHINE (SVM)

Prayoginingsih, Sila
Kusumawardani, Renny Pradina

SISFO
Vol 7 No 2 (2018)

This research performs classification on social media text, specifically for the case of customer complaint in the telecommunication industry. To represent complaint criteria relevant to telecommunication services, we use the categories used in myTelkomsel, a web application of Telkomsel. Although this application enables customers to file in their complaints directly in a self-service manner, many customers opt to post their complaints in the social media such as Twitter. Therefore, in this research we create a classification model using Support Vector Machines (SVMs) to enable the automatic categorization of such customer complaints. As the input for the training and testing process, we crawl Twitter using the Streaming API. The data is then filtered to get tweets containing information, complaints, criticisms, suggestions, and questions about Telkomsel’s products or services. Using RBF kernels optimized with grid search, the resulting classifier gives good accuracy and f-measure of 84.84% and 84.88%, respectively.