| Summary: | 碩士 === 國立中興大學 === 資訊管理學系所 === 98 === Online auctioning is one of the most successful network applications, and the price decision is one of the most important issues for both buyers and sellers. Although numerous support systems have been proposed for price predicting, all of them are base on a fixed prediction model. In the ever-changing market, the fixed model won’t be accurate all the time. This study proposes an updatable support system for price predicting of online auctions, and uses the combination of actual and simulated data to examine the performance. In our experiments, the system with a proposed updating scheme is compared with the traditional systems that without an updating process. The results show that the non-updating system become inaccurate as the time goes by, while the system with an updating scheme can still be accurate.
While online auctions have brought many benefits, it has also led to some potential dangers. Reputation systems are important because that it can distinguish between honest and dishonest participants. However, the existing systems (e.g. the systems used in eBay and Yahoo! Auction) are too simple that it cannot accurately reflect users’ trusts. In order to improve the existing schemes, this study proposes a novel reputation system that it takes the trading amount and the rater’s reputation into account. It treats seller reputation and buyer reputation respectively, as well as it provides users incentives to be honest over time. It’s robust to errors and easy to implement. The simulation result shows that it has outperformed the other existing schemes.
Reputation systems have become an important component of e- commerce. However, in the current systems, users can form a malicious group to unlawfully increase their reputation by deceiving the “feedback purchase”. In this study I propose a recommendation system aiming to find collusive users. Whenever a malicious user is identified (such as users blacklisted in the reputation center), we can find the suspected collusive users using the proposed scheme. The suspected colluders are sent to the reputation center to further investigate whether they are malicious users or not.
|
|---|
