Bank Customer Churn Predictor

Architecture

Description + stats

• Bank customer churn prediction application utilizing:

Name of model	Accuracy
Decision Tree	79.13%
K-Nearest Neighbors (KNN)	82.00%
Naive Bayes	82.25%
Random Forest Classifier	83.75%
Support Vector Machine (SVM)	84.13%
XGBoost Classifier	84.25%
XGBoost + SMOTE Classifier	83.87%
Voting Classifier	83.63%
Mistral Saba 24B LLM [OpenAI]	—

• It ingests 4000 entries to predict churn risk with visual insights, AI-generated explanations and emails.

Tech Stack

Purpose	Technologies
Core Tech
Frontend & Framework
Backend + DB
Other Libraries

Database + authentication

DB_Backend_Demo.mp4

Quick Start

1. Clone repo

2. pip install -r requirements.txt
   

3. Store below in a secrets.toml file under a .streamlit folder :

GROQ_API_KEY = ""
SUPABASE_URL = ""
SUPABASE_SERVICE_ROLE_KEY= ""
EMAILJS_PUBLIC_KEY= ""
EMAILJS_TEMPLATE_ID= ""
EMAILJS_SERVICE_ID= ""

4. streamlit run main.py
   

Research references + custom dataset badge-links

License

This project is licensed under the MIT License.