Gigi Auction Accurately Predicts Winning Bids Using Artificial Intelligence

[Asia Economy Reporter Onyu Lim] Gigi Auction has obtained a patent for a system that more accurately predicts the winning bid price of individual auction items.

Gigi Auction announced on the 15th that its "Winning Bid Price Prediction System and Method in Real Estate Auctions Using Genetic Algorithms" was recently registered as a patent. This system is characterized by improving the accuracy of winning bid price predictions by combining the existing statistical methods with genetic algorithm techniques, one of the artificial intelligence (AI) analysis methods.

Real estate auctions fundamentally require not only legal knowledge but also mathematical knowledge. This is because achieving the final goal of winning the bid requires precisely analyzing rights relationships and break-even points to calculate an appropriate bid price. To this end, Gigi Auction structured the system into four stages: 'data selection → multiple winning bid price predictions → comparison of winning bid prices → selection of the most accurate method.'

First, after inputting a total of 33 variables that affect the winning bid price in various fields such as winning bid rate, presence of tenants, and loan interest rates, data is selected. Then, winning bid prices are predicted and compared using three techniques: genetic algorithms, regression analysis, and artificial neural networks. Testing with Seoul’s overall annual winning bid rates (2013?2017) confirmed that the genetic algorithm technique’s accuracy was closest to the actual results.

After the first prediction comparison is completed, the final winning bid price is calculated through additional analysis. In this process, how to subdivide auction items by region is crucial to further improve accuracy. To this end, Gigi Auction subdivided regions based on appraised value and performed cluster analysis, then conducted a second genetic algorithm prediction for each cluster region to enhance accuracy. According to analysis subdividing Seoul into 25 districts, prediction accuracy improved only when classified by appraised value among appraised value, sale price, and winning bid price.

Meanwhile, since the 10th, Gigi Auction has been providing additional information such as certified copy of the register, tenant move-in history, and unpaid management fees for 'institutional sales' items independently conducted by each institution through Onbid.