Duke-Datafest2025

Team TLDR | ASA DataFest @ Duke University 2025
Contributors: Thomas You, Kevin Tao, Jason Deng, Sophia Shen, Lucy Wang

Project Summary

Where should I locate my new office?
Our project addresses this critical real estate decision using a combination of leasing transaction data, geospatial visualization, and machine learning. By analyzing leasing activity across US counties and industries, we built a model to support informed site selection—highlighting fast-growing sectors and evaluating top leasing factors across space, time, and location.

Project Structure

├── AddressConvertionTool/          # Tools for converting address-level data
├── ASA-Duke-Datafest...PDF         # DataFest prompt and analysis guide
├── Dataset/                        # Cleaned transactional datasets
├── GeoCoding Supporting Files/     # US county list and geocoding support
├── Geocoded Data/                  # County-level geocoded outputs
├── TemporalMapGifs/                # Animated gifs visualizing growth
├── Visualizations/                 # Static charts and spatial plots
├── county_scores/                  # County-based scoring results
├── .idea, .DS_Store                # IDE configs (safe to ignore)
├── DataFest 2025.pptx              # Final project presentation
├── DotPng.ipynb                    # Industry vs Rent analysis
├── MainLab_EDA.ipynb               # Exploratory Data Analysis notebook
├── MainLab_correlation.ipynb       # Correlation analysis
├── MainModel_Industry.ipynb        # Main predictive model notebook
├── MainModel_Transaction.ipynb     # Transaction type classification model

Methodology Highlights

• Temporal Analysis: Identified growing industries by year and county using animated maps and treemaps.
• Feature Engineering: Extracted top 10 features impacting leasing decision (e.g., rent, space, availability).
• Modeling: Developed a Decision Tree classifier with an accuracy of 92.5% for lease type prediction.
• Key Insight: Miami-Dade County is a hotspot for financial services, with high leasing activity and strong rental growth.

Tools Used

• Python (Pandas, Scikit-learn, Plotly, Folium)
• Jupyter Notebooks
• Geopy + US Census Geodata
• PowerPoint (final deliverable)

Visual Highlights

Animated Growth of Leasing Activity in Each County

Treemap of Leasing Activity Based on Industries

Random Forest Model Used for Classification

Recognition

Best Visualization Award at ASA DataFest 2025, hosted by Duke University.

License

This repository is intended for academic demonstration and is not currently licensed for commercial use.