자이랜드는 925만 호의 아파트와 223만 호의 빌라에 대한 주거용 AI 가치평가 서비스 베타테스트를 시작합니다 (한국시간 2022년 2월 3일 기준)

XAI Land is preparing to launch a beta test of its residential Automated Valuation Model (AVM) service with its banking and real estate partners.

What is being tested?

XAI Land’s residential AVM service is a service that provides accurate and unbiased valuations for 9.25 million apartments (아파트) and 2.23 million villas (연립/다세대주택) in South Korea.

Partners will test XAI Land’s valuations against their own internal back-data to assess the accuracy of XAI Land’s valuations. 

Who is it for (Use cases)?

1. Loan / mortgage & lending companies / banks – when mortgage applicants require a loan to buy a home a loan officer must understand the property’s value before assessing how much financing to provide the loan applicant.

2. Real estate brokers – brokers need a sense of a property’s market value before meeting a property owner who is looking to sell their property so they can assess whether or not the owner may have a reasonable expectation for the price of their property given market realities.

3. Real estate developers / investors – developers should determine an appropriate budget for property acquisitions when planning a new development where there are multiple home/property owners.

How does it work?

1. Web service – XAI Land will provide beta users access to our website link that can be accessed on any web browser in addition to login credentials (ID and password) that can be used to login.

Upon logging in, beta testers may then enter the address of the residential unit(s) they may wish to know the valuation of. A list of all residential units will appear alongside their respective valuations. 

2. API service – XAI Land has a GraphQL API that can be requested for as many valuations users may wish to request. 

How does XAI Land make its own valuations and ensure they are accurate? 

First, XAI Land collects data about South Korea’s various types of real estate including residential(villas, apartments, officetels, detached homes)/commercial (buildings, retail/office spaces, etc) uses, land parcels, and real estate transactions from the Ministry of Land, Infrastructure, and Transport (MOLIT). XAI Land also collects data from various public and private sources that can be used to characterize the built environment. For example, locations/density/type of retail/commercial establishments, locations of transportation infrastructure such as roads and mass transit, and location/types of other infrastructure and public amenities such as schools, hospitals, water features. 

XAI Land then prepares these separate data-sets to be analyzed and modelled by cleaning the data; these processes may include geocoding, merging, getting rid of outliers, getting rid of data that might be duplicated or fixing data-sets that might otherwise be incomplete. Once the data has been cleaned then the individual pieces of collected data are merged together to form one comprehensive data-set that can then be analyzed as a whole. 

Next, XAI Land trains AI-based models that analyze the comprehensive data-set to determine a property valuation based on a sales comparison / market-based appraisal approach.  

Finally, XAI Land tests its AI-based models by comparing valuation estimates that are produced by the model with real transactions from MOLIT that are not seen by the model during the training step. 

A final evaluation of the model is reflected when we calculate two key metrics – MdAPE (Median Absolute Percentage Error) and the Likelihood of a Valuation with an Error Rate of +/-10% and 20%, respectively. A sample of how we calculate these metrics can be viewed in the Excel below: