One-page case study
Financial Crisis Analysis Tool
Interactive financial analysis dashboard built at UMBC. Integrates FinBERT sentiment analysis on real-time financial news with Monte Carlo portfolio simulations and risk metrics (VaR, Sharpe ratio, volatility) across S&P 500 sectors and historical crisis periods (2008, COVID-19).
Proof Points
Monte Carlo risk modeling
FinBERT sentiment layer
Sector-level crisis analysis
Challenges
- • FinBERT model bias on domain-specific financial language required careful prompt and threshold tuning
- • Balancing technical risk metrics (VaR, drawdown) with interpretability for non-expert users
- • Optimizing Monte Carlo simulation runtime for large portfolios without sacrificing statistical accuracy
Learnings
- • Transformer-based NLP (FinBERT) for domain-specific sentiment classification
- • Monte Carlo simulation for portfolio VaR estimation at 95% confidence
- • GICS sector segmentation and cross-sector correlation analysis
- • Integrating market data (yFinance) with NLP signals for combined risk scoring
Stack
PythonFinBERTyFinancePandasNumPyScikit-learnPlotly