Table of Contents
ChurnGuard is designed to help banks lock in customer loyalty. The platform offers insights into a bank’s customer retention & the factors behind their churning. Then, given a customer being considered by bank employee & some basic information about the customer, we can predict the probability of churn and classify whether the customer is going to churn in the near future.
ChurnGuard was a project made by college students as part of a FinHackathon at UTDallas. To learn more about the winners, please look at our team. To learn more about the hackathon you can visit UTDallas Jindal News Center
Product Offerings:
- Historical customer analysis
- Customer churn prediction for bank customers
- LLM generated insights for result interpretation and reccomendations to retain customers
ChurnGuard predicts data using popular ensemble models (logistic regression, random forest, ADA boost, XG boost). The model results for each of ensemble models are below.
| Model | Accuracy | F1 | AUC | Specificity | Sensitivity |
|---|---|---|---|---|---|
| Random Forest | 83.14% | 59.23% | 0.7541 | 88.08% | 62.74% |
| ADA Boost | 73.14% | 45.12% | 0.6686 | 77.16% | 56.55% |
| XG Boost | 81.44% | 56.19% | 0.7368 | 86.41% | 60.97% |
| Logistic Regression | 59.22% | 37.52% | 0.6055 | 58.36% | 62.74% |
The project was built with the following libraries.
This is an example of how you may give instructions on setting up your project locally. To get a local copy up and running follow these simple example steps.
Please install the libraries listed in the requirements.txt file
- npm
conda install <library name>
- Get an API Key at https://openai.com/ for ChatGPT
- Clone the repo
git clone https://github.com/harshak913/ChurnGuard
- Install Python Packages from requirements.txt
conda install <library name>
- Enter your API as an environment variable called
OPENAI_API_KEYopenai_api_key = 'ENTER YOUR API'
- Run the streamlit library
streamlit run test_dashboard.py
To make a prediction, use the website to enter the values for the customer in the text input section. Run the program by pressing the submit button to generate new prediciton for customer.
- Improve UI for OpenAI responses
- Add chat bot feature to allow managers to ask questions about model interpretation
- Finetune underlying models for predictions using a wider hyperparameter space
- Host website on live webserver
- Add data about bank to LLM to create better reccommendations to managers
Distributed under the MIT License.
- Harsha Gurram - LinkedIn
- Harsha Kolachina - LinkedIn
- Mihir Padsumbiya - LinkedIn
- Viswa Kotra - LinkedIn
Project Link: https://github.com/salaga-py2021/ChurnGuard
Use this space to list resources you find helpful and would like to give credit to. I've included a few of my favorites to kick things off!