Comparison of LSTM and GRU Methods in Sentiment Analysis of Satusehat Appication Review

Abstract

This article describes a comparison between the LSTM and GRU methods in sentiment analysis. Both methods were chosen based on the simplicity of the model and are considered capable of processing relatively long data. the dataset used in this study amounted to 12260 data obtained from scraping on the Google Play site using google-play-scraper library. In analyzing Sentiments, 2 Deep Learning methods are used, namely LSTM and GRU. From the results The GRU accuracy value reaches 91% while the LSTM accuracy value is 89%. However, for the Recall and F1-score tests, these two methods still get low scores in analyzing positive data.