Abstract
On-line auction becomes more and more popular now. Many people engage in developing the mechanism that makes on-line auction safety and effective. The learning effect is an important impression in the repeated on-line auction. The phenomenon over-bidding will be brought by the dynamic evaluation from the learning effect in the repeated on-line auction. There are many factors influencing the dynamic evaluation of the on-line auction. In this research, three parameters, number of bidders, number of blocks and market price of product are investigated to influence the results including average payoff over price, over-bidding frequency and over-bidding rate. We use artificial agent instead of the participators in the auction and simulation analysis with genetic algorithm to observe and analyze the results. We find some interesting trends in the repeated on-line auction. (1) When the number of bidders increases, the average payoff / price decreases, the over-bidding frequency increases, and the over-bidding rate decreases. It is due to the more number of bidders, the more competitive they are. (2) When the market price of product increases, the average payoff / price increases, the over-bidding frequency decreases, and the over-bidding rate decreases. It is due to facing of higher price product, bidders will be more cautious to bid. (3) When the number of blocks increases, the average payoff / price is uncertain, the over-bidding frequency increases, and the over-bidding rate increases. It is due to the learning effect of the bidders and the mutation of genetic algorithm. Thus it can be seen, the learning effect affect the auction result indeed.