Cognitive Real Estate Assessment (CREA)

Use case

Private sellers and buyers, real estate professionals, banks, insurance companies, investment funds and other stakeholders oftentimes depend heavily on the accurate value estimation of their real estate properties. Yet, finding the right price is a complex and daunting task, which requires expert knowledge, considerable amounts of research, and numerous comparisons with other offers on the market. Furthermore, expert appraisals can be quite expensive and, even though they are thorough and usually based on extensive data, they rarely turn out to be accurate.

Solution

In cooperation with several real estate professionals, Cognexa created an AI-powered solution using a wide range of different methods in machine learning, financial mathematics and mathematical modelling. Trained on extensive databases of real estate advertisements and transactions, CREA offers a pricing model based on a regression core. Utilizing the knowledge of the relationship between price and other parameters of the property, the calibration is strengthened by a post-processing pipeline, which models the relationship between price levels in geographically neighboring areas.

In pre-processing, the important building blocks are:

• method for price adjustment – historic vs. current prices
• process of dealing with the unknown – missing values of parameters in advertisement and transaction databases

Result

CREA is available on www.CMN.sk and allows users to calculate property values and plot price level maps. While it is still in its pilot phase, CREA is already outperforming other competing services, such as IFP apartment price calculator, and is capable of pricing apartments in Slovakia with a median relative error below 8%. To prove its validity, independent testing against actual transactional data further confirmed CREA’s unique accuracy of value estimation.