Indonesian Digital Wallet Sentiment Analysis Using CNN And LSTM Method

In this paper, we present a deep learning based approach to sentiment analysis on Indonesian digital wallet review. We collected the reviews of the following digital wallets: DANA, OVO, Link Aja, and Sakuku by obtaining the data from Google Play Store. Sentiment analysis from comments on Google Play Store is used to extract user opinions on various aspects: accessibility, transaction, account, service, and performance. We applied the CNN LSTM architecture on performing the sentiment analysis. The experimental results show that the training accuracy is 87% and the validation accuracy is 83%. Based on the data and analysis results of the four digital wallet applications, namely DANA, OVO, LinkAja, and Sakuku, the sentiments tend to be negative towards the categories we have created, including transactions and access, services, accounts, and performance.