以社會網路與電影本體為架構之電影推薦系統

A Movie Recommender System based on Social Networks and Movie Ontology

本研究提出結合社會網路與本體論架構之電影推薦系統，期望能透過社會網路和電影本體的整合，加強人與人、項目與項目、及人與項目之間多維度的關係，藉以輔助電影推薦服務的完整性。本研究建立電影本體，以清楚地表示各電影類型的階層屬性，藉此瞭解不同類型電影之問的關係，並利用電影類型階層和電影特徵值之概念做為推論相似電影的依據。本研究以5點式Likert問卷方式了解使用者對此系統之滿意程度，有效樣本為42份。分析結果顯示大部分的使用者對於本系統之推薦服務皆感到滿意。此外，本研究亦運用查全率（Recall）與平均絕對誤差（Mean Absolute Error, MAE）比較本系統推薦方法與傳統使用者平均評價方法兩者推薦的效能。其結果亦顯示本系統推薦方法優於傳統的平均評價方法。

This paper presents a movie recommendation system which integrates social networks with ontology technologies. The aim of this study is to improve the integrity for movie recommendation services by taking account of the relationships among people, association among items, as well as relationships between people and items. The establishment of movie ontology not only identified users' movie preferences and attributes with higher accuracy, but also improve the overall performance of the recommendation system. Total 42 adults were participated in this study and completed the questionnaires. The result showed that most of participants were satisfied with the recommendation services from the system. This study also employed two measures, Recall and MAE (Mean Absolute Error), to compare the proposed approach in the system with the traditional users' average ratings. The experimental result showed that the efficacy of the proposed approach was better than that in the traditional users' average ratings.