Analisis Sentimen Ulasan Aplikasi Info BMKG di Google Play Menggunakan TF-IDF dan Support Vector Machine

Abstract

Posting online reviews has become one of the most popular ways to express opinions and sentiments towards service applications. The Meteorology, Climatology and Geophysics Agency (BMKG) Info application is an Android and iOS-based mobile application that provides information on weather, climate, air quality, and earthquakes that occur in various regions in Indonesia. The information contained in this application is very important but has a worse value than other forecasting applications. Sentiment analysis is the process of classifying text into several classes such as positive sentiment, negative or not containing both. This research aims to analyze user reviews on the BMKG Info application from the Google Play website. The benefits obtained are as consideration for developers to improve the shortcomings of the application. The classification process uses Term Frequency-Inverse Document Frequency (TF-IDF) and the Support Vector Machine (SVM) algorithm. This research successfully collected 2500 reviews from users of the BMKG Info application on the Google PlayStore website using the web scraping method. Text preprocessing of the reviews used case folding, symbolic and stopword removal, tokenization, normalization, and stemming. User ratings help in identifying the sentiment label of a review, 66% of reviews are positive while the rest are negative. The most frequently reviewed topics with sentiment value are "application", "information", "update". This research conducted three experimental scenarios based on the composition of training data and test data. Based on the prediction model, the scenario with 75%:25% split data has the highest accuracy rate of 79%.