XYZ Bank Customer Churn Predictor
by Soumil-mukho
About this product
• Spearheaded the development of a robust customer churn application capable of processing over 4,000 customer records, utilizing Python for backend logic, Supabase for scalable database management, and EmailJS for automated communication workflows. Achieved peak classification accuracy using Support Vector Machines (SVM) at 84.13% and XGBoost at 84.25%, following extensive model experimentation and validation. Integrated SMOTE (Synthetic Minority Over-sampling Technique) to address class imbalance, significantly improving model generalization and fairness across minority segments. Hyperparameter tuning was conducted using grid search and cross-validation to optimize performance metrics including precision, recall, and F1-score.
• Engineered a domain-specific feature pipeline tailored for banking churn prediction, embedding business logic to extract behavioral patterns and transaction anomalies. This pipeline powered the Qwen3 32B LLM to generate personalized churn insights, enabling banks to understand customer attrition risks at a granular level. Delivered these insights through interactive dashboards built with Streamlit, enriched with dynamic visualizations using Plotly. Additionally, the system auto-drafted retention emails using HTML, CSS, and JavaScript, allowing real-time intervention strategies to be deployed by relationship managers.
• Designed a modular full-stack machine learning architecture integrating Groq API for ultra-fast inference, Supabase for real-time data syncing, and Streamlit for seamless UI/UX. Analytical components were built using SciPy and Scikit-Learn, while visualization and reporting were handled via Plotly. The project also leveraged a suite of essential libraries including NumPy, Pandas, Utils, OS, Base64, Re, Pillow, and DateTime for data preprocessing, image handling, regex parsing, and secure encoding. The entire system was optimized for scalability, interpretability, and automation, enabling financial institutions to proactively mitigate churn through data-driven decision-making. This end-to-end solution exemplifies the fusion of machine learning, LLMs, and full-stack engineering to deliver actionable intelligence in high-stakes environments.
Demo Login Details
sdf12@gmail.com
Password
Asdg#1357
Changelog
New Feature
- Version 2.0.0
Improvement
- 1)Added state & session management, ensuring that, even on page refreshes after login, the page stays in current state without reverting back to original state & getting logged out, overriding default behaviour. 2)Hiding topbar.
Standard License
There is no additional charge for updates. You can continue to download the latest version for free.
-
Single end product Use in one project only
-
Personal or commercial use Suitable for business projects
-
No redistribution Cannot be resold or shared
Refer & Earn 10% New
Share with friends and earn commission on their purchases.
Soumil-mukho
