Recently Zillow announced that the qualifying round of Zillow® Prize, the machine learning competition with a $1 million grand prize, has closed following record-breaking participation from scientific minds around the globe. More than 4,350 participants from 78 countries submitted over 76,000 contest entries, averaging 517 a day, in an attempt to improve the Zestimate® home valuation accuracy.
Zillow Prize is the second largest competition ever hosted on Kaggle and to date, ranks as one of the top three most popular machine learning contests ever, according to the data competition platform.
“We’ve been blown away by the engagement of people from around the world in the Zillow Prize,” said Dr. Stan Humphries, Zillow Group chief analytics officer. “That fact that so many individuals from around the world are excited by this competition really speaks to both the importance of understanding a home’s value to consumers and the intellectual challenge in getting it right. We’ve worked hard to improve the Zestimate algorithm’s accuracy rate to 4.3 percent nationwide, and now we’re eager to see the innovation within the thousands of submissions shared by the global data science community.”
With the close of the qualifying round, all submissions will be scored against the actual sales prices of homes sold over the next three months. On January 17, 2018, Zillow will announce the top 100 teams who have earned their spot in the final round to compete for the $1-million-dollar grand prize.
In the final round, competitors must build an algorithm to predict the actual sale price itself and will be allowed to use outside data not provided by Zillow. To walk away with the $1-million-dollar grand prize, the winning algorithm must beat Zillow’s benchmark accuracy on the final round competition data seti and be more accurate than any other competitor’s algorithm.
Zillow publishes Zestimates on more than 100 million homes across the country based on 7.5 million statistical and machine learning models that examine hundreds of data points on each individual home. To calculate the Zestimate home valuation, Zillow uses data from county recorder and tax assessor records, and direct feeds from hundreds of multiple listing services and brokerages, and information entered by homeowners.
The Zestimate home valuation is a designed to serve as starting point to understanding the value of a home and is not an appraisal. Homeowners have the ability to update facts about their homes in order to impact to their Zestimate. More than 70 million homes on Zillow have been updated by the community of users.
i Final round participants are challenged to beat the Zillow benchmark model, a modified version of the Zestimate algorithm that will be trained using the exact same data set available to everyone in the final round. This benchmark model has been created for the purposed of this competition and is different than the standard Zestimate displayed on the website.