Deep Learning for Sentiment Analysis to Predict the Probability of Bank Loan Default

Abstract

Social networks have taken the world by storm with their fast and commendable speed. It could be social, political, or present with all sorts of situations that arise. People’s opinions around the globe are articulated through social media, making it apposite for drawing out opinions. Organizations that aim at refining their products and services use sentimental analysis methods to increase their resources. In the banking and financial industry, it is much easier to get feedback from customers through Twitter and or Facebook sentimental analysis. The elements associated with Twitter or consumers and services providers who want to know who they are, and what they are in their daily life towards their bank and financial portfolios cannot suppress Facebook sentimental analysis. Hence, this study aims to predict the probability of bank loan default and classify the Twitter messages by exhibiting the results of deep learning algorithms. High-performance computing with hyper-parameter space for grid-search (HPSGS) and hyper-parameter optimization (HPO) are developed and compared with the effectiveness of three gradient boosting decision trees. The results reveal that the XGboot algorithm has a better prediction or features a score that is better as compared to other algorithms at 91 percent in the test data and 93 percent performance in the validation data. It is also seen that women are more likely to default than men as across all the algorithms, their likelihood of risk or default is higher than that of men. These results are useful for decision-makers and the financial sector for future use and planning in credit risk and bank loan default-prone areas.