Optimasi Textblob Menggunakan Support Vector Machine Untuk Analisis Sentimen (Studi Kasus Layanan Telkomsel)

Abstract

Sentiment analysis refers to Natural Language Processing techniques that are classified as Unsupervised Learning to identify positive, negative, or neutral opinions. Many of these opinions come through Twitter, because social media is quite effective and efficient in commenting because it can only write a maximum of 140 characters. From previous research, the value of the accuracy of the sentiment analysis carried out by one of the NLP libraries, namely TextBlob, has shown that Unsupervised Learning does not produce such good scores. With the Telkomsel service case study the writer took the dataset from Twitter and the results of the analysis with TextBlob only showed a value of 58.59%. Optimization is done by adding the Support Vector Machine method which is included in the Supervised Learning category. The best results obtained from this study are values that show 75%.