SOCIAL MEDIA ANALYSIS FOR INVESTIGATING CONSUMER SENTIMENT ON MOBILE BANKING

• Main Author: Giovanna Asali, Alessandra
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/61217
• Institution: Institut Teknologi Bandung

Life nowadays is inseparable from the presence of internet, accessible seamlessly through a handheld device. One thing that has also become very popular through the presence of internet is various methods of cashless payment through financial technology. The growth of fintech adoption is constantly increasing every year, including Indonesia. Commercial banks usually have their own version of fintech called mobile banking. Due to COVID-19 pandemic and people’s fear of making physical contact, mobile banking usage grows even more. Seeing the current trend of everything online, banks need to keep up with their mobile banking service to satisfy their customer and stay competitive in the market. However, failure to understand customers will result in costly, ineffective development. This paper is aimed to give more insight for banks to understand consumers’ sentiment in social media towards their mobile banking service, with the objectives of discovering consumer’s sentiment regarding mobile banking service before and after COVID-19 pandemic, finding out the most and least favored features, determining area(s) that requires improvement, and giving recommendation where refinement is due. Nine mobile banking features are assessed, determined through manual observation on Twitter, namely: payment, block, open new bank account, login, transaction report, bank balance, top-up, transaction, and transfer. Twitter entries from January 1st, 2019 to December 31st, 2020 which include the words ‘mobile banking’ and its variations plus one of the nine features are collected through a text mining process. January 2019 until February 2020 is considered a condition before COVID-19 pandemic and March 2020 until December is considered a condition after COVID-19. 30% of data are labeled with their sentiment manually to be used as training data. Labeled data are then applied on two text classification method: Neural Network and Naïve Bayes. Neural network model shows better accuracy at 79.8% compared to Naïve Bayes with only 57.5%, and therefore is used for the research. For the whole data, negative tweets have the biggest proportion at 49.8%, then followed by neutral tweets with 44.5% and positive tweets with 5.7% only. It is found that average number of total tweets per month decreases from before to after COVID-19 breakout, negative tweets per month decreases, while neutral and positive tweets increase. To make a binary classification for each of the nine mobile banking features, neutral tweets are removed. It is found that the proportion of negative tweets are way higher than the positive tweets for all categories, with transaction feature receiving the least negative tweets at 79.04% and login feature receiving the most negative tweets at 100%. Going more in depth to uncover the problems in each feature, words with the highest frequencies and some representative example tweets are listed. In conclusion, although average total tweets before and after COVID- 19 is decreasing, the average for positive tweets shows growth. This indicates growing user satisfaction overtime. Based on the proportion of negative and positive tweets, most to least favored features in mobile banking are transaction, open new account, payment, transaction report, block, bank balance, top up, and finally login. Since all features garnered more negative sentiment compared to the positive ones, it can be concluded that all nine assessed features require refinement. Some recommendations that can be given to banks to improve their mobile banking service performance are no extra fee policy to access any feature, providing follow up or record that can be recalled after every important action or transaction, making sure that the mobile banking app stays updated to customer needs while also keeping it steady, fast, and easy-to use, and better synchronization especially in transactions with third party entities such as other banks and e-money and digital wallet service provider.