結合社群之協同過濾推薦系統

SoLoMo-based Collaborative Filtering Recommendation System

黃榕萱(Rong-Syuan Huang)

指導教授 ： 段裘慶

國立臺北科技大學/電資學院/電腦與通訊研究所/碩士(2013年)

摘要

在智慧型手機使用人數快速成長的現在，適地性服務(LBS)成為一個極為重要的功能。但在眾多的LBS推薦系統中，如何盡可能的反映用戶的喜好，並推薦合適的項目，才是最為重要的。以往此類推薦系統皆以用戶的歷史評價與大眾的興趣排行作為推薦依據，但實際而言，會主動評分的用戶為少數，造成系統無法準確推薦合適的項目。本論文提出結合社群之協同過濾推薦系統(SCF)，利用社群不同程度的分享功能，作為一興趣景點(POI)評分參考標的，如此能大幅改善顯性評分不足之情況，並善加利用用戶的好友群，從中尋找與用戶興趣相投之使用者，如此之推薦項目更能投用戶之喜好，更能精確推薦適合的POI給用戶，且因系統不再漫無目的的所有用戶做興趣比對，因此能減少系統之資料運算時間。
根據模擬實驗結果，本論文所提出之SCF相較於LCFDTP、DFBT與TWCF策略，不僅平均推薦時間與推薦誤差減少，也有較高之推薦覆蓋率。在推薦誤差方面，在用戶速度為50 km/hr，查詢範圍為0.5 km時，SCF之推薦誤差平均優於其他策略約58 %。在平均推薦時間方面，平均優於其他策略約49 %。最後，在推薦覆蓋率方面，平均優於其他策略約7 %。因此本論文所提出之SCF更能達到優良的推薦效果。

關鍵字

適地性 ； 打卡 ； 協同過濾 ； 社群

並列摘要

In the rapid growth in smart phone users, location-based services (LBS) have become an extremely important service. However, in many LBS recommendation systems, the most important things is how to accurately response the user's preferences, and recommend some suitable items. In the past, such a recommendation system was based on the historical evaluation of the user and everyone interested ranking. However, only a small number of users ever take the initiative to score items, causing the system cannot accurately recommend suitable items. In this paper, we proposed SoLoMo-based Collaborative Filtering Recommendation System. Using the social website's check-in as a point of interest (POI) score, it thus can significantly improve the deficiencies explicit rating, and more precise recommended suitable POIs to the users.
The simulation result showed that SCF has not only less recommend error and more recommend coverage, but also less average recommend time than LCFDTP, DFBT and TWCF. In recommend error, when user’s velocity is 50 km/hr and query range is 0.5 km, SCF is better than the others about 58 %. In average recommend time, SCF is better than the others about 49 %. Finally, in recommend coverage, SCF is better than the others about 7 %. These showed that SCF has better effect on recommendation.

並列關鍵字

LBS ； Check-in ； Collaborative Filtering ； SoLoMo

參考文獻

19. G. Linden, B. Smith and J. York, “Amazon.com recommendations: item-to-item collaborative filtering,” IEEE Internet Computing 7, vol. 7, Issue 1, 2003, pp. 76-80.

25. S. Gong, “A collaborative filtering recommendation algorithm based on user clustering and item clustering,” Journal of Software, vol. 5, no. 7, 2010, pp.745-752.

36. D. John, “Do data characteristics change according to the number of scale points used? an experiment using 5-point, 7-point and 10-point scales,” International Journal of Market Research, vol.50, no. 1, 2008, pp. 61-77.

46. Y. Koren, R. Bell and C. Volinsky, “Matrix factorization techniques for recommender systems,” IEEE Computer, vol. 42, Issue 8, 2009, pp. 30-37.

47. R. M. Bell and Y. Koren, “Scalable collaborative filtering with jointly derived neighborhood interpolation weights,” in Proc. of the 7th IEEE International Conference on Data Mining (ICDM 2007), Omaha, USA, Oct., 2007, pp. 43-52.