Cleaning of Auction Data for Bidding Decision

• Main Authors: Shu-gang Han, 韓樹岡
• Other Authors: 張嘉惠
• Format: Others
• Language: en_US
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/80700885818321388897

碩士 === 國立中央大學 === 資訊工程研究所 === 95 === Bidding for products on the Internet has become a common activity in our daily life. However, it’s a tedious problem that there are too many items for the bidder to select the cheapest one. In the results providing by eBay, only a small number of results are target items. This is a common situation while the user is searching for a main product in 3C. We aim at helping the bidder compare items easily on auction websites. In this thesis we propose CADBid, which is a web-based system built between auction websites and the bidder. CADBid is able to automatically filter out non-target items and clean the descriptions about these items. Afterward, a list is generated which helps the bidder compare these items. The list only shows the target items along with their important properties. Our work focuses on two tasks. The first task is item filtering. The second is cleaning of descriptions. After cleaning of descriptions, the clean descriptions are used to assist the first task. We view the two tasks as classification problems and propose two feature sets. We build two classification models based on Support Vector Model. Our experiment shows that cleaning of description is helpful because clean descriptions indeed improve the accuracy of item filtering. With CADBid, the bidder will be convenient while making a good decision on which item to bid.