San Francisco Apartments
I have friends who are entering the apartment renting market, and family members asking for the right rates to charge tenants in multi-family apartments.
What it does
This app scrapes popular apartment websites, and collects the data for a major city near you. It then takes that data and carefully cleans it into a form that it can be digested using Google's Prediction API. Using thousands of listings with associated information (size, location, and rating), it helps determine the market standard price for any apartment based on the characteristics you choose, so that you can always find the best deal.
How I built it
I used the webscraper.io along with a mouse recorder to spider and scrape information from some of the most popular apartment renting websites. I then aggregated these data for 3 major cities (D.C., N.Y., and S.F.), and cleaned the data into a form where it could be read in using Google's Prediction API. Using regression based machine learning, I then had an application that could output a price based on selected input information, depending on the user (such as the number of bedrooms).
Challenges I ran into
It was surprisingly difficult to clean the data into a usable form, as even the smallest deformations in the data set could choke the process.
Accomplishments that I'm proud of
I'm proud that I was able to determine a reasonable system for determining whether or not friends that used the app might be able to determine good value for the apartments that might be rented.
What I learned
I learned more about machine learning, the importance of well organized information, the increasing abundance of such information to the public, the capability of Google's Prediction API, and more about ruby.
What's next for apartmentPrediction
The incorporation of workplace commute information to help determine the value of certain locations in a city. For example, someone might find a place to be more valuable if it is closer to their place of work.