Sentiment Analysis of Bank BRI Services using Word Embedding and Gated Recurrent Unit Method

This title has been presented on Thursday, December 14, 2023 at 15.50-16.00 GMT+7.

Abstract

This title has been presented on Thursday, December 14, 2023 at 15.50-16.00 GMT+7.

The widespread use of social media has significantly influenced various aspects of community life, including the way suggestions and complaints are communicated. Bank BRI is one institution that utilizes Twitter as a platform to gather suggestions and complaints from its customers. However, technological advancements have also brought about new challenges, particularly in promptly acquiring customer feedback. This challenge can be addressed through sentiment analysis, an automated text data processing method aimed at obtaining information regarding the sentiments conveyed within public opinions. By analyzing sentiment in public postings, conclusions can be drawn about whether the services of Bank Rakyat Indonesia tend to elicit more positive or negative responses. This analysis employs the deep learning method known as Gated Recurrent Unit (GRU), complemented by Word Embedding word2vec applied to 1200 Indonesian public Twitter data independently crawled. The classification outcomes of GRU using word2vec demonstrate an accuracy of 0.7520, precision of 0.7546, recall of 0.7572, and an F1-score of 0.7556. This performance proves superior compared to the scenario without the use of word2vec, which yielded an accuracy of 0.7032, precision of 0.6914, recall of 0.7032, and an F1-score of 0.6899.