← Back to Portfolio
Florida Real Estate Investment Analysis
What I Built: A machine learning model using 10 years of Florida government data (634 GB) to predict single-family home resale profitability.
View on GitHub |
Watch Video Demo
Key Skills Demonstrated
- Large-scale data engineering (634 GB, 700+ files)
- SQL database design & optimization
- Machine learning (Logistic Regression, Random Forest)
- Government data experience (FDOR)
- Handling imbalanced datasets (SMOTE technique)
What I Delivered
Business Problem: Predict which properties will be profitable when resold within the same year.
Technical Approach:
- Data Engineering: Automated processing of 700 files; built SQL Server database from scratch.
- Analysis: SQL queries to identify trends and track iBuyer activity (OpenDoor, Offerpad).
- Modeling: Random Forest model with high recall for identifying loss properties.
- Key Finding: Geographic location (subdivision risk score) is the strongest profitability predictor.
Business Value:
An actionable risk assessment framework for real estate investment decisions.
Why This Matters
- Government data expertise (FDOR)
- SQL for data engineering & analysis
- Predictive modeling & forecasting
- Outputs ready for Power BI visualization
- Workflow mirrors JIRA-managed processes