This paper presents a method to automatically group financial consumer complaints by the type of product they are about, using Natural Language Processing (NLP). The dataset was cleaned and prepared by removing unnecessary words, breaking it into tokens, and converting it into numbers using TF-IDF. Then, Machine learning models are trained to predict which product each complaint is related to. Among the models tested, Stochastic Gradient Descent gave good results. The findings show that NLP can help financial institutions handle complaints faster and more accurately by sorting them automatically into product categories. This approach can be helpful for banks and regulators in improving how they respond to customer issues.
Mutesi et al. (Tue,) studied this question.