-
Consumers can face challenges with financial products and services, leading to complaints that may not always be resolved directly with financial institutions. The Consumer Financial Protection Bureau (CFPB) acts as a mediator in these scenarios. However, consumers often struggle to categorize their complaints accurately, leading to inefficiencies in the resolution process.
-
Our project aims to facilitate faster complaint submission and resolution by automatically categorizing complaints directly based on narrative descriptions, enhancing the efficiency of complaint management and smoothly routing it to the appropriate teams.
-
We developed a Hybrid approach by leveraging language models (BERT) and traditional machine learning techniques to find a trade-off between computational complexity and the need for model re-training. We've fine-tuned DistilBERT using the Consumer Complaints data on the main product and issue categories, and used supervised classifiers for categorizing sub-products and sub-issues.
The implementation of our project has two primary impacts:
-
Ease for Consumers: Automates the tagging of complaints into appropriate categories, reducing the need for consumers to understand complex financial product categories.
-
Industry Adoption: Offers a streamlined approach to complaint handling that can be adopted by financial institutions beyond the CFPB, promoting consistency across the industry.
Checkout the app deployed on HuggingFace Spaces : https://huggingface.co/spaces/Mahesh9/CFPB-Complaint-Classifier
1. Clone the Repository
git clone "https://github.com/mahesh973/TagMyComplaint.git"
2. Navigate to the Directory
cd "TagMyComplaint"
3. Install the necessary dependencies
pip install -r requirements.txt
4. Navigate to the app Directory
cd src
5. Launch the application
streamlit run main.py
After completing these steps, the application should be running on your local server. Open your web browser and navigate to http://localhost:8501 to start exploring the Consumer Complaint Insights 2023.
- This application is built using Streamlit.
- For more detailed information to explore the raw data, visit the official Consumer Complaints Database: CFPB Complaints Database
- Link to the data file we've used for training
Feel free to contribute to the project by submitting issues or pull requests on GitHub. Your feedback and contributions are highly appreciated!