摘要 網路拍賣是近年來最興盛的網路交易模式,隨著網路頻寬的增大,可上網的裝置越來越普及化,網路拍賣的市場也隨之逐年成長,拍賣的品項也是包羅萬象,也因為拍賣市場的大眾化,網拍詐騙的事件也層出不窮,如何防範網拍詐欺也成為網拍平台業者的所需面臨重要課題,目前各網拍平台業者也都採用評價機制來建立一套賣家與賣家的信任機制,希望有效的減少網拍詐欺事件的產生,進而影響到買賣雙方對交易平台的信任,本研究希望可以利用網拍平台所提供的交易資訊,利用社會網路分析,可以建構出網路拍賣異常帳號線上即時偵測系統,讓買賣雙方在原有的交易平台上的評價機制外,多一項可以檢驗對方交易狀況的資訊系統,避免與有異常的帳號進行交易。本研究從Yahoo! 拍賣自動擷取線上拍賣交易評價資料和拍賣物品資訊,並使用所提出之社會網路指標與關連性指標進行實作與驗證。資料取得方面設計一套Auction Crawler Services可以自動下載拍賣網站的所需資訊,利用proxy清單設定進行多線程資料擷取作業,定期對於所擷取的交易資訊使用J48進行資料探勘,找出異常探勘原則組,利用此原則組查詢判斷為異常交易帳號之可能。
Internet auction is the most popular trading model in recent years. With the enlarged internet bandwidth and the more and more prevailing internet access facilities, the internet auction market is also growing up year by year, and besides the auctioned items are all-inclusive. At the same time, for the popularity of auction market, the deceptions have also never stopped. Therefore, how to prevent the fraud of internet auction has confronted and tested the internet auction platform providers. Longing for the effectively reduced deception in internet auction, nowadays, the internet auction platform providers all adopt evaluation mechanism to build up a system of mutual trust mechanism between the sellers and buyers, and thus promoting the trusts of both sellers and buyers to the transaction platform. The purpose of this thesis is to construct a prompt online detection system for suspected abnormal account in internet auction through the Social Network Analysis by taking advantage of the transaction information offered by the platform providers. Besides of the currently existing evaluation mechanism of transaction platform, both sellers and buyers can have another information system in inspecting the transaction condition of opposite party, so as to prevent from the transaction with suspected account. In this thesis, the transaction evaluation information of online auction and the information about the auctioned products are retrieved from Yahoo! auction automatically, and then the proposed Social Network Analysis Measure and the Releated Measure are used for the experiment and verification. To get the information, a system of Auction Crawler Services is designed, which can automatically download all the required information from the auction website. Through the set of proxy check list, the multi-thread data acquisition is performed, and the acquired transaction information is investigated via the using of J48 Data mining, so as to find out a classification group of fraud investigation. The classification group then can be adopted to check and determine the possibility in deceit of transaction account.