在拍賣極為盛行的時代，許多公司早已使用拍賣的方式來取代傳統採購的行為，而當公司的採購決定以拍賣的方式解決，往往三方(下標者、買方或賣方、電子交易市集供者)都會面臨到相當大的壓力，拍賣的價差可能影響著公司的這筆生意是盈餘或虧損，但在拍賣結束前並無法確知最後價格。另外拍賣交易進行的時間也關係著三方人員需在網站上耗費的時間。因此實務上，三方人員都希望能盡快知道可能的拍賣結標價與所需時間。 本研究提出一個改良式屬性導向歸納演算法，從過去的拍賣資料建立出價時間與出價的概念樹並歸納出序列規則。這些規則能夠對未來結標價與結標價格分別作預測，並給予一個期望的百分比；最後驗證的結果發現，我們能夠對於結標時間作有效的預測，而對於結標價還有改善的空間。 In this thesis, we proposed a modified attribute oriented induction method to find rules which can predict close price (the actual price paid by winner) and the duration (from the beginning of auction to the last bid) of Business-to-Business auctions exercised on a commercial website. The uncertainty faced both sellers and buyers is an inherent feature of auctions. While a seller is anxious for knowing the actual selling price, a Bidder has very short time to decide to bid or not to bid at this moment, especially when a little difference may cause very big loss or profit to a company. Even the provider of the e-marketplace also cares about whether this is an efficient auction when the close price should be higher (lower) than the reserve price in a forward (backward) auction, how long the auction would be taken and when the employees should come back to handle the results. Therefore, how fast one can predict the outcome of the auction and how accurate the prediction is became a critical problem. We focused on two important variables of auction, time and price, and collected auction data of backward English auction format of three buyers which represent three distinct companies from a B-to-B auction e-marketplace provider in the past two years. The rules summarized from datasets by our modified attribute oriented induction can estimate the predicted close price and duration after several bids are taken when an auction opened
