Data-X Project - Realest8

Overview

Realest8 predicts pricing and price trends of properties in the San Francisco area to inform real estate investors' decision-making.

Link to Running Web-app

Includes a web-app and iPython Notebook that both have the following functionality:

• Predict Airbnb listing price based on general features
• Predict Airbnb listing price change in a year from now
• Visualize/inform the user of how many days the property would have to be rented out on Airbnb to match the average monthly price of regular rent in the area

Interactive Demo - Web-app

1. cd web
2. python app.py

Sample input:

• Longitude: -122.434
• Latitude: 37.747
• Accommodate: 2
• Bedrooms: 1
• Bathrooms: 1
• Beds: 1
• Room Type: Private Room
• Neighborhood: Noe Valley
• Zip Code: 94131

Assumes your machine has numpy, pandas, matplotlib, sklearn, and Flask.

Interactive Demo - iPython Notebook

1. cd ipython_notebooks
2. ipython notebook project_pipeline.ipynb
3. Go to the Kernel tab and click Restart and Run All
4. Follow the input prompts at the bottom.

Sample input:

• Longitude: -122.434
• Latitude: 37.747
• Accommodate: 2
• Bedrooms: 1
• Bathrooms: 1
• Beds: 1
• Room Type: Private Room
• Neighborhood: Noe Valley
• Zip Code: 94131

Assumes your machine has numpy, pandas, matplotlib, sklearn, and ipython notebook.

Repo Overview

• exported_models - .hdf and .pkl files of cleaned DataFrames and sklearn models. Used to optimize setup of ML models for project.
• ipython_notebooks
  • airbnb_price_predictor.ipynb
    • Cleans November 2018 Airbnb Dataset
    • Trains various regression models (while tuning hyper-parameters using cross-validation) to predict Airbnb listing price from general features
    • Chooses the best model with lowest Median Average Error
    • Exports plots, cleaned DataFrames, and best sklearn regression model
  • airbnb_price_trends.ipynb
    • Cleans November 2018 and other past Airbnb Datasets
    • Trains various regression models (while tuning hyper-parameters using cross-validation) to predict price trend in a year
    • Chooses the best model with lowest Median Average Error
    • Exports plots, cleaned DataFrames, and best sklearn regression model
  • airbnb_rent_comparison.ipynb
    • Uses cleaned data from airbnb_price_predictor.ipynb
    • Provides visualization functions to compare Airbnb prices to average monthly rent prices in the area
  • project_pipeline.ipynb
    • Uses cleaned data from airbnb_price_predictor.ipynb and airbnb_price_trends.ipynb
    • Uses visualization functions from airbnb_rent_comparison.ipynb
    • Sets up ML models for price prediction and price trend prediction
    • Provides methods to clean and format input data for ML models
    • Takes in user input to use to predict current Airbnb price and Airbnb price in a year
    • Visualizes comparison of Airbnb prices and average monthly rent prices in the area
• plots - saved plots from iPython Notebooks
• raw_datasets - .csv files of uncleaned data from dataset sources (sources in the README)
• web
  • app.py - Flask app
  • static - static CSS
  • templates - Jinja2 templates for pages

Datasets

• InsideAirbnb
  • November 2016 Listings
  • November 2017 Listings
  • May 2018 Listings
  • November 2018 Listings
• RentCafe
  • January 2017, May 2018, August 2018, November 2018
  • Independently scraped and formatted into .csv files

Libraries

• Flask
  • Jinja2 Templating
• numpy
• pandas
• matplotlib
• sklearn
• iPython Notebook

Credits

This project was a term-long project made for UC Berkeley's Data-X (IEOR 135) course.