標題: 以使用者喜好多樣性及跨領域分析為基礎之線上商品推薦方法
Online Product Recommendation Approach based on Diversity and Cross-domain Analysis of User Preferences
作者: 林俞君
劉敦仁
Lin, Yu-Chun
Liu, Duen-Ren
資訊管理研究所
關鍵字: 商品推薦;多樣性;矩陣分解;隱含主題模式;跨領域;Product Recommendation;Diversity;Latent Topic Model;Matrix Factorization;Cross-domain
公開日期: 2016
摘要: 互聯網之蓬勃發展促使新型態的網路(新聞)資訊與商務交易平台之興起,愈來愈多使用者透過網路資訊平台來獲取新聞或特定主題如旅遊、餐飲、娛樂、美妝、生活時尚等資訊,或是利用電子商務平台來消費購買商品。資訊平台之多角化發展,使得資訊平台提供之相關資訊量爆炸且愈趨複雜,因此,藉由良好的線上推薦機制來推薦使用者有興趣之相關資訊,並提高使用者點擊率與忠誠度,是網站資訊平台相關研究重要的議題。 本研究整合考量跨資訊來源喜好分析及喜好多樣性分析研發一個新的線上商品推薦機制,並進行線上推薦評估與比較。網站資訊平台對於商品購買而言,大部分使用者為cold-start 使用者,傳統推薦方法對於cold-start使用者之推薦成效較不佳。因此本研究研發設計以矩陣分解(MF)及隱含主題模式(LDA)探勘為基礎,來進行使用者新聞文章瀏覽與商品購買之跨領域喜好分析,預測使用者歷史與線上商品喜好,並分析其喜好多樣性或單一性程度,研發以使用者喜好多樣性程度為基礎之線上商品推薦方法,以提升cold-start使用者之推薦效果。實驗結果顯示,本研究所提出的方法能有效改善cold-start問題,並提升使用者商品點擊率。
The flourishing of the Internet has increasingly promoted the rise of new types of online news websites with e-commerce portals. Combining online news websites with e-commerce can attract more users and create more benefit, which is also an important trend of online worlds. The great amount of information provided by news websites is becoming even more complicated. Therefore, it is an important issue for online news websites to deploy appropriate online recommendation methods that can raise the users’ click-through rates and loyalty. In our research, we conducted cross-domain and diversity analysis of user preferences to develop novel online product recommendation methods, and evaluate online recommendation results. Accordingly, by cross-domain analysis on news browsing and product purchasing, we developed methods based on Matrix Factorization and Latent Dirichlet Allocation to predict the user preferences for products. Our experimental result shows that our proposed approach can improve the cold-start problem and enhance the click-through rate of products.
URI: http://etd.lib.nctu.edu.tw/cdrfb3/record/nctu/#GT070353407
http://hdl.handle.net/11536/138613
顯示於類別:畢業論文