Complaint Classification Using Support Vector Machine for Indonesian Text Dataset
Desi Ramayanti, Umniy Salamah · Zenodo (CERN European Organization for Nuclear Research) · 2018
Text classification is used to classify text data, for example, to find some information from news stories and text from social media that can be used by data owner. Since manual classification is time-consuming and difficult, many studies have been done to this research area. However, the most of studies focused on English text classification. This research attempted to classify Indonesian text dataset by using SVM classifiers. We have conducted research to classify Indonesian text using Python programming language and scikit-learn library. As the result, the experiment without cross validation and tuning parameter for SVM classifier on the dataset achieved the accuracy 0.89473 with value of precision and recall is 0.90289 and 0.89473 respectively. Moreover, value of K for SVM classifier is 0.78992 so that strength of agreement is included into good category. Then, the experiment using cross validation with k-5 and k-10 and tuning parameter with C constant and gamma value. Result of cross validation with k-10 is derived the best accuracy with value 0.9648, however, it spend computational time as long as 40.118 second. Then, we conducted experiment to find the best kernel function among Sigmoid, Linear and RBF. Moreover, based on result of experiment, kernel function Sigmoid achieved the best accuracy and computational time.