Improved Algorithm for Home Valuations

Published Mar-08-19

A sophisticated algorithm to improve an online tool for home valuation.

Zillow Group, United States

The Story:

Improved Algorithm for Home Valuations Zillow is an online real estate database that has an algorithm called a Zestimate that predicts the estimated market value of a home. It is automatically computed daily with public and user-submitted data, taking into account special features, location and market conditions.

The company claims that its calculator is wrong in home price predictions about 4.5 percent of the time. In an attempt to get that error rate below 4 percent it launched a global open innovation contest called the Zillow Data Prize on the Kaggle platform. Up for grabs was a prize of $1 million.

The contest was structured in two rounds, the initial public qualifying round and the private round for the top 100 qualifying teams. The first round attracted nearly 4,000 teams and individual competitors from 91 countries. They were tasked with developing an algorithm to make predictions about the future sales prices of homes with the aim of improving the Zestimate residual error. Those who made it through this stage then had to build a home valuation algorithm from the ground up.

Finalists spent about six months working on their algorithms. The overall winner was ChaNJestimate, a three-person team that formed midway through the competition with members hailing from Morocco, Canada and the United States. They were Chahhou Mohamed, Nima Shahbazi and Jordan Meyer.

Improved Accuracy

Nima and Chanhou entered as a two-person team and after advancing to the final round persuaded Jordan to join their group. This allowed them to make an even more sophisticated algorithm that blended their separate models from the first round. Their winning algorithm combines several machine learning techniques including deep neural networks and turned out to be 13.3 percent more accurate than Zillow’s benchmark model.

“People are incredibly passionate about their home and understanding its value, and we are amazed by the winning team’s hard work the past two years to make the Zestimate even more precise,” said Stan Humphries, chief analytics officer and creator of the Zestimate, in a news release.

“We’ve been on a 13-year journey making the Zestimate more accurate, and hosting Zillow Prize allowed us to invite thousands of brilliant data scientists from around the world to join us on this journey. We’re so proud that the winning team’s huge achievement, and the work of all the teams in the competition, will provide millions of homeowners with a better understanding of one of their biggest life investments.”

Updated Model

Following the open innovation data competition, Zillow’s data scientists began incorporating the winning team’s work into the Zestimate model. They claim that the winning algorithm will help push Zestimate’s current nationwide error rate of 4.5 percent to below 4 percent.

Share on      
Next Story »

What Can we Solve for You?