A complaint classification system
The following project is a datathon project where the problem statement and dataset was provided by BNY Mellon. Problem statement: Develop an intelligent system that aims to leverage natural language processing techniques to develop an efficient and accurate system for bank customer complaint analysis, classification. By automating the classification, financial institutions can enhance their dispute resolution capabilities and improve customer satisfaction.
The dataset consists of 2 columns - Product and Narratives, in which Product contains the different types of complaints: credit card, mortgages and loans, debt collection, retail banking and credit reporting. The dataset has 162420 rows in total. The aim was to classify the complaints into the 5 categories accurately.
Sentiment Analysis on the dataset has also been performed in order to identify how the user is feeling while submitting a complaint and a chatbot was created to interact with the user while they submit a complaint and relay their query to management. The model was further integrated with a front-end via streamlit.