Implementasi Bidirectional LSTM untuk Analisis Sentimen Terhadap Layanan Grab Indonesia

Abstract

Grab Indonesia is one of the leading online motorcycle taxi companies in Indonesia and has a large number of customers in Indonesia. The level of customer satisfaction varies with the services provided, so there must be suggestions and complaints from customers. Sentiment analysis can be used as a solution to determine the level of service satisfaction in order to improve the system and service. This study aims to determine the level of satisfaction of Grab Indonesia users through the Grab application in the Playstore. One of the approaches that can be used is LSTM. LSTM is an RNN algorithm development to solve the vanishing gradient problem. LSTM has the disadvantage of only running can only capture information from one direction. Bidirectional LSTM (BiLSTM) is an LSTM method that has been developed, where BiLSTM can capture information from two directions. In this BiLSTM method, the more data, the better the algorithm's performance. The test results show that BiLSTM is more reliable than LSTM in the case of sentiment analysis on the Indonesian Grab service. BiLSTM produces the best accuracy of 91% and training loss of 28%. Suggestions for future research can produce more and varied word representations by considering the word embedding combinations.