Altos Research has launched an automated valuation model that uses current real estate listings to forecast future housing valuation trends.
The AltosEvaluate Forward Valuation Model uses the same “machine learning” technology that credit card companies use to notify borrowers of suspicious transactions. The algorithms review the borrower’s transaction history, along with the transaction history of every other borrower in the dataset, to determine when a purchase may be fraudulent.
“This math and software is really sophisticated and for the first time we can really apply it to this vast local market database that we have to give our clients a forecast and a view at the local level of how home valuations are going to change,” Altos Research CEO Michael Simonsen told National Mortgage News.
“In the mortgage markets, whether they’re bond or whole loan traders, or originators, anybody who has exposure, they can use that information to understand things like loan-to-value, to say, ‘Are these things heading up or down and what do I do financially?’” Simonsen added.
Other mortgage applications include the real estate owned industry, where asset managers and real estate brokers can use the technology to make decisions on disposing of distressed assets from foreclosures.
The FVM combines the machine learning technology with Altos Research’s database of active real estate listings in 20,000 ZIP codes nationwide. Simonsen said that information provides a better view of the current market than datasets based on closed transactions that take months to complete and can be infrequent in slower-moving markets. That data gives the FVM a leg up on other predictive modeling technology because it’s timely and localized, allowing future valuations of up to 12 months, Simonsen said.
“There’s so much signal in the active market, so why wait until the recorded transaction,” he said.
