Gan, Kim Soon (2009) A comparative study for parameter selection in online auctions. Masters thesis, Univerisiti Malaysia Sabah.
In this information-rich age, online auctions have become an important procurement tool in either commercial or personal use. As the number of auctions increases, the process of monitoring, tracking bid and bidding in multiple auctions become a problem. The user needs to monitor many auctions sites, picks the right auction to participate, and makes the right bid in making sure that the desired item satisfies the user's preference. All these tasks are somewhat complex and time consuming. The task even gets more complicated when there are different start and end times and when the auctions employ different protocols. Due to the complex and dynamic nature of the online auction, one of the strategies employed is using genetic algorithm to discover the best strategy. Hence, this work attempts to improve an existing bidding strategy by taking into accounts the evolution of various model of generic algorithm in optimizing the parameter of the bidding strategies. In this work, three different models of genetic algorithms are considered. In the first model, the crossover and the mutation rate of the genetic algorithms are varied in order to create different combination of crossover and mutation rate. The new combination of genetic probabilities from this investigation will eventually perform better than the recommended genetic probabilities adopted in the previous work. The second model is the dynamic adaptation model namely the dynamic deterministic adaptive model. The bidding strategy from the experimental result of this experiment will eventually perform better than the bidding strategy that applied fixed static genetic operator's probabilities. Self adaptation genetic algorithm is the last model that will be used to evolve the bidding strategy. The bidding strategies applying self-adaptation model are expected to perform better than the deterministic dynamic adaptation because of the nature of the algorithm itself. The evaluations are conducted in a simulated online auction framework with multiple auctions running concurrently. The effectiveness of the bidding strategies is measured based on the average fitness of the individuals, the success rate and average payoff in obtaining the item in the auctions. The performance of these bidding strategies will be empirically demonstrated in this thesis.
|Item Type:||Thesis (Masters)|
|Uncontrolled Keywords:||E-commerce, Online auction, Generic algorithm, Tracking bid, Bidding|
|Subjects:||?? HF5469.7-5481 ??|
|Divisions:||SCHOOL > School of Engineering and Information Technology|
|Deposited By:||IR Admin|
|Deposited On:||13 Feb 2012 11:03|
|Last Modified:||13 Feb 2012 11:03|
Repository Staff Only: item control page