(第 26 期)   第十三卷第二期   2019 年 6 月 30 日出刊

強化電影類型偏好之互動式推薦方法

Interactive Movie Recommendations with Strengthened Analysis of Genre Preferences

本文關鍵字:層級分析法導演定錨程序類型定錨程序偏好穩定度推薦系統 Analytic Hierarchy Process( AHP)Director-based anchoringGenre-based anchoringPreference stability

本文摘要

推薦技術在電子商務網站可挖掘潛在商業價值並扮演重要角色,過去廣受歡迎協同式過濾推薦系統能透過找到與使用者偏好相似的使用者,進而推薦與興趣相符合的商品,但協同式過濾不能克服冷啟動和評分稀疏的問題。本研究著眼於此,以電影推薦為研究對象,提出有別於以往混合過濾( hybrid filtering)方法,透過使用者對電影類型( genre)與導演進行定錨程序以量測使用者類型偏好穩定度,提供更準確的個人化推薦結果。近期研究指出使用者的偏好穩定性將影響他們的決策過程,特別是經驗性商品;因此,是否能透過量測使用者的偏好穩定度提供有效推薦是值得探討的議題。具體來說,研究以關聯規則分析電影類型關聯,改善評分稀疏問題並透過電影類型與導演定錨程序強化計算使用者的電影類型偏好。本研究採用層級分析法( Analytic Hierarchy Process,簡稱 AHP),透過一系列成對比較進行電影類型與導演定錨程序推薦符合使用者偏好的電影。實驗結果顯示經過類型合併後能有效的改善使用者與電影評分數據稀疏性的問題,且經過導演強化類型定錨過程後能提升推薦準確度,特別是偏好穩定度低的使用者,本研究結果將可提供電子商務推薦網站參考。

E-commerce often relies upon buyer recommendation techniques given their potential commercial value as well as recommendation systems’ ability to accurately predict user interests. For these systems, collaborative filtering (CF) enables websites to recommend products for target users based on the preferences of peers with similar interests. While CF can expand a user’s profile of interests, it cannot overcome the problems of cold start and sparse ratings. Since recent studies have shown that the stability of users’ preferences influences their decision making, especially concerning experiential goods, measuring such stability regarding these goods is worth investigating. This study thus proposes integrating genre- and director-based anchoring processes to identify users’ preferences for movie genres and measure preference stability in order to provide more precisely personalized recommendations. Specifically, we overcome the problem of sparse ratings by analyzing associations among movie genres and directors. By employing the analytic hierarchy process (AHP), we thus infer user preferences for movie genres via a series of interactive genre- and director-based anchoring processes, which ultimately provides effective, precisely interactive movie recommendations. The experimental results show that the system can tackle the sparse rating problem by merging similar genres and then can increase recommendation ability by strengthening analysis of genre preferences via directors, especially for users with unstable preferences for movie genres. The research results can serve as a reference for e-store business.
本文附件:

本期同分類其他文章


本刊著作權屬於「中華民國圖書館學會」所有。
Powered By Vanilla Journal - 香草期刊系統 0.256 / 2006 - 2007 © Weizhong Yang