📅 Period: 2000–2024
🧠 Tech Stack: Python | Power BI | SQL | Pandas | Scikit-learn
This project analyzes patterns and predicts business failures in the U.S. using FDIC Failed Bank data and BLS Business Employment Dynamics (Business Survival Rates).
The goal is to support data-driven insights for regulators, investors, and policymakers to mitigate financial risks and enhance economic resilience.
- FDIC Failed Bank List
Contains historical bank closure data, including location, acquisition details, and failure causes. - BLS Business Employment Dynamics – Business Survival Rates
Tracks establishment openings, closures, and survival percentages by year and sector.
- Data Preparation – Cleaned and merged FDIC and BLS datasets.
- Exploratory Data Analysis – Identified trends in bank failures and business survival rates.
- Predictive Modeling – Applied Logistic Regression, Random Forest, and XGBoost for failure prediction.
- Visualization – Designed a Power BI dashboard highlighting state-wise failures, trends, and survival patterns.
- Businesses in finance, manufacturing, and retail sectors show distinct survival curves.
- Strong correlation between macroeconomic factors and industry failure rates.
- Predictive accuracy of 91% achieved using XGBoost classifier.
- Power BI visuals reveal recovery clusters and geographic concentration of failed institutions.
Figure 1: U.S. bank failure trends and survival patterns across states.
Figure 2: Georgia, Florida, and Illinois recorded the highest cumulative bank failures during the 2008–2012 period.
Developed a reliable AI-driven business failure prediction framework integrating government datasets.
Supports U.S. regulators, investors, and policymakers in identifying high-risk sectors early and reinforcing financial stability.
Dipon Das Rahul
🎓 MBA in Business Analytics (STEM), Midwestern State University
📍 Texas, USA
📧 dipondasrahul@gmail.com
🔗 LinkedIn | GitHub
#AI #MachineLearning #FinancialRisk #BusinessFailure #FDIC #BLS #Python #PowerBI