透過您的圖書館登入
IP:184.72.135.210
  • 學位論文

考慮商品異質性改善協同過濾推薦系統

Exploiting Item Heterogeneity for Collaborative Filtering Recommendation

指導教授 : 魏志平 王俊程
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


在今日電子商務與知識經濟盛行的環境中,網路上大量流通的資訊成為我們主要 的資訊來源管道,但資訊使用者卻也面臨了資訊超載的挑戰,因此必須尋求可以 協助過濾與挑選商品的篩選機制。同時,線上商家為了提高顧客滿意度與忠誠度, 也需要可以協助加強顧客關係維護的機制。推薦系統便是因應資訊超載與加強顧 客關係維護而產生的一種解決之道,有了推薦系統的服務,資訊提供者便可以適 時地提供合適的產品給使用者,同時使用者也可以免去因面臨過多商品資訊而無 法做出決策的困境。在眾多已發展的推薦技術中,協同過濾推薦機制 (collaborative filtering approach)是其中最成功且最廣受採用的推薦技術, 但是這項推薦技術卻因為沒有考慮到受推薦的商品彼此間的差異性,亦即是當計 算使用者相似度以及進行使用者喜好預測時,所有的商品評分都被視為是同等重 要的,如此一來可能導致參考不可靠的使用者意見,以致無法達到理想的推薦效 果。因此,在本研究中,我們提出一個考慮個別商品內容差異性的協同過濾推薦 技術(item-based content-weighted collaborative filtering approach),藉 以解決傳統協同過濾技術不考慮商品異質性的缺點。實驗結果顯示本研究所提出 的方法的確可以在考慮內容差異性的情況下,提供比其他兩種協同過濾方式更高 的預測準確度(prediction accuracy),同時,為了進一步提升可預測的涵蓋範 圍(coverage),我們也提出一個結合傳統協同過濾技術的混合方法(hybrid approach),其實驗結果也顯示所提出的混合方法,可以達到與傳統協同過濾技 術相同的可預測的涵蓋範圍,而仍然有較佳的推薦效果。

並列摘要


In today’s electronic commerce and knowledge economy environments, information users may experience information overload and seek for e-services to help them in filtering and selecting from an overwhelming array of product information. Online merchandisers may also seek to better manage customer relationships that lead to higher customer satisfaction and loyalty. In response, recommendation systems have emerged as a class of e-service that are not only address the challenge of information overload by suggesting products of greatest interest to users, but also facilitates organizations better managing their customer relationships. Among various recommendation techniques, the collaborative filtering approach is the most successful and widely adopted one. However, the basic design of traditional collaborative filtering approach ignores item heterogeneities during the recommendation process. That is, all the user preferences on items are deemed identically important and given an equal weight in measuring user similarities and predicting user preferences. This may take unreliable users into consideration and therefore, lead to poor recommendations. In this research, we design an item-based content-weighted collaborative filtering approach to improve the recommendation effectiveness by considering the content similarities of items. We conduct a serious of experiments to evaluate our proposed approach. The empirical evaluation demonstrates that our proposed approach can achieve better prediction accuracy than those of the benchmarks. To further promote the prediction coverage of our proposed approach, we also create a hybrid approach by combing our proposed approach with the traditional one. The corresponding experimental results also show that the hybrid approach can achieve the same coverage as the traditional one and its prediction accuracy is still better than that of the traditional one.

參考文獻


Wei, C., Chiang, R., Tseng, C., and Yang, C., “Web Recommendation Services: A
National Conference on Artificial Intelligence (AAAI-98), 1998, pp.714-720.
Berson, A. and Smith, S. J., Data Warehousing, Data Mining & OLAP, McGraw-Hill,
Inc., 1997.
Algorithms for Collaborative Filtering,” Proceedings of the 14th Annual

延伸閱讀