電子商務推薦平台基於社群標籤與行為模式

自Web 2.0的概念萌發開始，任何網路使用者皆能透過網路上傳分享檔案，也因為如此網路上的資源也以飛躍般的速度成長增加，而標籤(Tag)的使用也開始受到重視，藉由幾個關鍵字(Key Words)或一些詞彙(Terms)來描述檔案或文件的內容，增加了網路資源的再用性以及可檢索性，現在更衍生出了以群眾分類法(Folksonomy)概念為基礎的社群標籤(Social Tag)，改變以往只有管理者群才能對資源進行標籤的動作，讓使用者彼此的標籤能互相共享，提高了網路資源分類的準確性及可靠性。藉由標籤的廣泛使用，也讓我們對於使用者檔案的個人化更加容易，透過紀錄使用者在網路上的行為模式(Behavior Pattern)，將其瀏覽或互動過的資源之標籤進行收集並整理成標籤雲(Tag Cloud)，可從此觀察出該使用者較對某類型事物較有興趣或是專注在某個領域上，因此本文提出了一個電子商務推薦平台，並將其建立在分散式系統上，透過使用者在網頁瀏覽商品或是進行標籤、交易等動作，系統將自動建立使用者的個人檔案(User Profile)，並針對不同的使用者行為模式來調整使用者檔案內各標籤權重以達到個人化的目標，最後透過餘弦相似度(Cosine Similarity)對使用者個人化檔案及商品檔案進行比對，找出與使用者檔案相似度較高、較符合使用者喜好的商品來進行推薦，增加商品的搜索性及分類準確性，並發掘潛藏的商機，改善以往只使用關鍵字檢索造成推薦結果不盡如人意的情況。

關鍵字

社群標籤；群眾分類法；個人檔案；行為模式；餘弦相似度

並列摘要

The development of Web 2.0 has enabled all internet users to upload and share files through the internet. It is also because of this that the amount of resources on the internet has grown rapidly. An increasing degree of importance is being attached to the use of tags, in which a few keywords or some terms describes the content of a file or document. This greatly enhances the reusability and searchability of web resources. Social tags, a derivative based on the concept of folksonomy, changed the convention in which only administrators can tag resources, allowing users to share their tags and increasing the accuracy and reliability of web resource classification. The wide use of tags has also simplified the personalization of user profiles. By recording the behavioral patterns of users on the internet, collecting the resource tags in their browsing history or interactions, and creating tag clouds with the tags, one can understand what matters a user is interested in or what domains they focus on. We therefore developed an e-commerce recommendation platform on a distributed file system. And automatically creates user profiles based on products the users have browsed, tags they have made, and transactions they have engaged in. Based on the behavioral pattern of the user, the weights of the tags in their user profiles are adjusted to achieve the objective of personalization. Finally, we adopted cosine similarity to compare user profiles with product files so that products that closely matched the preferences of the user could be recommended. This enhances the searchability of the product and the accuracy of the classification. Moreover, it can excavate possible business opportunities and rectify the previous situation in which poor recommendations were derived from keyword searches.

並列關鍵字

social tag ； folksonomy ； user profile ； tag cloud ； cosine similarity

參考文獻

[2] Jesse Vig, 2010, “Intelligent Tagging Interfaces: Beyond Folksonomy”, UIST’10

[4] Hak-Lae Kim, John G. Breslin, Sung-Kwon Yang and Hong-Gee Kim, 2008, “Social Se-mantic Cloud of Tag: Semantic Model for Social Tagging”, KES-AMSTA 2008, LNAI 4953, pp. 83–92

[5] David Vallet, Ivan Cantador, Joemon M. Jose, 2010, “Personalizing Web Search with Folksonomy-based User and Document Profiles”, In Advances in Information Retrieval, Vol. 5993, pp. 420-431

[12] Xiaogang Han, Zhiqi Shen, Chunyan Miao, Xudong Luo, 2010, “Folksonomy-Based Ontological User Interest Profile Modeling and Its Application in Personalized Search”, AMT 2010, LNCS 6335, pp. 34-46

[13] Martin Szomszor, Ivan Cantador and Harith Alani, 2008, “Correlating User Profiles from Multiple Folksonomies”, HT’08

國際替代計量

電子商務推薦平台基於社群標籤與行為模式

未授權

主題瀏覽