Uses Zillow metadata, NLP on realtor description, and VGG16 on home images to predict home sale prices in Portland from 6/16 - 7/17.
Switch branches/tags
Nothing to show
Clone or download
Latest commit 337dfc1 Sep 10, 2017
Permalink
Failed to load latest commit information.
README.md
analyze_data.ipynb first commit Sep 10, 2017
clean_home_data.ipynb first commit Sep 10, 2017
get_home_data.ipynb first commit Sep 10, 2017
predict_home_prices.ipynb first commit Sep 10, 2017
single_family_homes_one_hot.csv

README.md

The get_home_data file scrapes Portland homes sold between 7/16 and 7/17 from the Portland Maps API. It then scrapes Zillow metadata and Redfin images on those houses.

clean_home_data cleans up the scraped data.

analze_data creates a word vectorization of the realtor description and performs NLP sentiment analysis on it. It also uses VGG16 to get features from the home pictures. It combines the Zillow metadata and this other info into a matrix and then uses Gradient Boosting to predict home prices.

predict_home_prices explores some of the predictions in the test set.