Predicting the satisfaction level of mobile banking users of Bangladesh from social media sites using machine learning

Abstract

The research focuses on the prediction of customer satisfaction in the mobile banking industry in Bangladesh, using social media data collected with Google Forms. The 2608- entry dataset categorizes target attributes as Non_Satisfied or Satisfied. The research carefully analyzes the performance of numerous machine learning models, including Bernoulli Naive Bayes, Support Vector Machine, Logistic Regression, K-Nearest Neighbours, Decision Tree, Long Short-Term Memory (LSTM), and Convolutional Neural Network (CNN).Actually, the LSTM deep learning model matches others, collecting an excellent 99.82% accuracy. This high accuracy shows its capacity to model changes in time within social media data, providing a deeper knowledge of issues changing customer satisfaction in the particular case of mobile banking in Bangladesh.The dataset, which was collected from Google Forms, provides an extensive variety of user opinions and offers a strong foundation for training and figuring out the models. The results show how important it is to use advanced deep learning methods, especially LSTM, to find complex patterns in social media data and make accurate predictions. The effects reach mobile banking service providers as well, providing useful tips to improve customer satisfaction and experience.Finally, this study shows the useful use of LSTM to improve mobile banking services based on users' pointed out views using social media platforms, providing useful data specifically modified to the Bangladeshi market. The important accuracy achieved by LSTM highlights its practical uses to improve and modify mobile banking services.