Portfolio
Data science and analytics projects: economics, machine learning, and data visualization through hands-on analysis.
Economics & Finance
Predicting Next Month's Unemployment Rate
Sentiment and language features from FOMC meeting minutes predict the next month's unemployment rate through NLP-driven feature engineering
- Python
- Streamlit
- scikit-learn
- NLP
Predicting Median Real Estate Prices in Select US Cities
SARIMA models with automatic parameter selection forecast median home prices across 7 U.S. metros, validated by walk-forward backtesting
- Python
- Streamlit
- statsmodels
- SARIMA
Retail Trade Time Series Forecasting
SARIMA model forecasts retail trade volumes within 1.8-2.6% error for Seattle and Dallas metro areas
- Python
- statsmodels
- SARIMA
- Pandas
PPP Loan Program Exploratory Data Analysis
89% of PPP loans went to businesses over 2 years old, with April 2020 seeing the highest disbursement spike
- Python
- Power BI
- Pandas
Conference Board Survey Data Analysis
India had the highest average industrial production, 35% higher than the UK which ranked second
- Python
- Power BI
- Pandas
Data Visualization
Economic Tracker Dashboard
Real-time county-level tracking reveals uneven economic recovery across U.S. metro areas post-pandemic
- Python
- Streamlit
- Pandas
- Plotly
Real Estate Heatmap Dashboard
Interactive dashboard analyzing 1,262 neighborhoods across 7 U.S. metros with heatmaps, rankings, and market velocity analysis
- Python
- Pandas
- Streamlit
- Plotly
YouTube Statistics Analysis
Music channels generate disproportionately more views per upload than any other YouTube category
- Python
- scikit-learn
- Matplotlib
- Pandas
Machine Learning
NFL Big Data Bowl
Tackle acceleration is the strongest predictor of successful NFL tackles, more important than closing speed
- Python
- scikit-learn
- Pandas
- Matplotlib
Student Dropout Prediction
CatBoost model predicts student dropout with 89% accuracy, identifying marital status and nationality as strongest risk factors
- Python
- CatBoost
- SMOTE
- scikit-learn
Predicting Credit Default (Lending Club)
Random Forest achieved 69% AUC on credit default prediction, highlighting the difficulty of consumer credit modeling
- R
- Logistic Regression
- LASSO
- Random Forest
Smartphone Models Insights
RAM capacity, not price, is the strongest predictor of smartphone user ratings
- Python
- CatBoost
- Matplotlib
- Pandas
Mathematics & Statistics
The Secretary Problem
37% is the optimal stopping point - interview a third of candidates to establish a baseline, then hire the next one who exceeds it
- Python
- NumPy
- Matplotlib