建構B2B電子商務推薦機制之研究

Abstract

企業為了強化其整體競爭優勢，在B2B電子商務運作上，透過個人化推薦系統，可以讓企業縮短商品或物料搜尋的次數和時間，得到符合其需求的產品資訊，以達到快速交貨的目標。本研究針對B2B的交易資料進行分析，以訂單資料庫及產品資料庫為基礎，透過資料探勘技術與推薦機制，建構一個能夠依照使用者訂單行為給予適當項目推薦的B2B電子商務系統，以達到個人化推薦。本研究之推薦機制主要是利用客戶過去的交易行為做客戶分群，並針對客戶群進行推薦，以合作式過濾推薦系統為基礎，提供四種推薦模組並實作B2B電子商務推薦系統。本研究建構一個人化的動態網站環境，以提供客戶一個專屬的差異化資訊服務。最後，本研究也對系統所提供的四個推薦模式進行實驗評估、分析，以驗證與比較四種推薦模式之推薦精確度。

To consolidate the competitive advantages of enterprises in B2B e-commerce operation, the personalized recommender system may help companies to shorten the searching time of products and materials and to obtain the product information which satisfies their needs for delivering products quickly. This research uses the transaction data of B2B, order and product database, data mining techniques and recommendation methods to build a B2B e-commerce system which personally recommends suitable items to users according to their order behavior. This recommendation mechanism groups customer into many clusters based on their transaction behavior and then uses collaborative recommendation method to recommend items. Four kinds of recommendation module are proposed for implementing our B2B e-commerce recommender system. In this study, our system can provide differential information service to customers in a personalized dynamic web environment. Finally, the experiments are conducted to verify and compare these four recommendation models.