Zillow Hackathon: HopCity
HopCity has developed a neighborhood matchmaking technology that uses machine learning from dynamic, publicly available data to better match people to a new neighborhood before they move. To help over 40 million people moving every year, HopCity asks the user's current location to recommend neighborhoods in the city they would like to move. The matching algorithm shows the top neighborhoods based on a machine-learned combination of various characteristics, and can be further refined via filters interesting to the user, e.g. age, income, attractions, restaurants, nightlife, safety, etc.
The application was created on February 6-8th, 2015. You can view it live here.
Video Demo: http://youtu.be/vVrgYxbyRvw
Challenge and Approach
Our submission is for Challenge #3: helping first-time homebuyers. How do I find a home that meets my needs, within my budget, in an area that fits my priorities?
Our approach for solving this problem for 8 million people per year who move interstate and almost 40 million people who move between cities was to:
- alleviate the fear of moving to a new city by matching them with lifestyle & community neighborhood insights in a unique and easily visual way.
- crunch available data sets and match homebuyers with neighborhoods that are affordable and a close match to their price point.
- enrich the decision-making process by presenting relevant information pulled from variety of source like: Freebase Knowledge Graph, Yelp, Zillow, Flickr and etc
- target millenials - the largest first-home buyer generation in the next 10 years by using the Yelp API to show more visual images and descriptions.
- match the user, once they settle on a neighborhood, with a set of housing options (Homebuyers then can go to Zillow listings in this area)
- Andrey Nokhrin @fliptrealestate
- Sam Rexford @samrexford
- Chris Li www.linkedin.com/in/chrili/en
- Derik Strattan www.derikstrattan.com
- Evgeny Podlepaev www.linkedin.com/in/podlepaev/en
Technologies, APIs, and Datasets Utilized
We made use of:
- Zillow Neighborhood data API which conveniently combines real estate, demographic and other insights to access neighborhood characteristics
- Machine Learning including Latent Semantic Analysis, Clustering and Vector Space Model to implement a simplified prototype of an unsupervised Recommendation algorithm tested empirically
- Flickr API to display neighborhood images that are dynamic, like the nature of these communities!
- Yelp API to display cafes, restaurants, nightclubs, gyms, parks, theaters and markets - all the attributes important to selecting a neighborhood. We show the top results for each category
- Google Images API to pull in images for each neighborhood on the results page
- Freebase Knowledge Graph API to pull additional structural data about neighborhoods like description, etc.
- IPython Notebook contains some code for data clustering and dimensionality reduction
In order to build & run our app:
sudo easy_install pip virtualenv virtualenv venv source venv/bin/activate pip install -r requirements.txt
Our code is licensed under the MIT License. Pull requests will be accepted to this repo, pending review and approval.