以網站探勘技術為基礎之電子目錄上推薦服務之研究

Abstract

隨著電子商務的發展，不論是B2B或者是B2C的商業模式，電子目錄儼然已經成為買賣雙方的主要介面。電子目錄不再只是提供商品的規格，更是提供客戶服務的重要媒介。然而，一個擁有豐富資訊的網站，如何對於不熟悉網站架構的使用者，或對購買商品特性不熟悉的使用者，協助他們做出採購的決策呢？本文試圖透過網站探勘技術，瞭解使用者的瀏覽目的，並將此探勘所得到的瀏覽樣式，推薦給具有相同需求的使用者做為參考。同時，將推薦分為同類商品、相關產品、以及其他產品資訊說明三種方式推薦，讓使用者更清楚推薦原因。本文以網站探勘的技術為基礎，提出概念限制型參考(CCR, Conceptual Constrained Reference)的交易識別方式，來確認交易為某類責訊的瀏覽，藉此確認使用者目的。之復，利用改良的混合型順序性(MIXSEQ)相似度比對，對瀏覽路徑做叢集(Clustering)處理，以做為推薦的資料集。在推薦策略方面，先將網頁做分類，區分為內容型網頁與導覽型網頁，推薦時以內容為主要的推薦網頁，以提高推薦實用性。

As the E-Commerce prevalence, E-catalog has become an important interface to a company through which customers interact, regardless of B2B or B2C. E-catalog not only provides product information, but becomes the important media of service providing for customers. However, how does an E-catalog assist casual/unfamiliar users with such rich information in their purchasing decision making? The challenge for Websites is how we know users needs? Thus, we will apply Web Mining technology to understand user`s needs and get usage pattern from previous browsing experience. In this paper, we propose a new transaction identification, Conceptual-Constrained Reference (CCR), for understanding the goal of users in browsing the e-catalog. Furthermore, we use the MIXSEQ similarity measure for web-usage clustering to create the recommendation dataset. Finally, we will classify pages into two types-content pages and navigation pages, and recommend those pages which are content types for users when they are browsing.