結合情境資訊與社群之動態時段協同過濾推薦系統

Dynamic Time Periods Collaborative Filtering Recommendation System based on Contextual Information and Social Network

吳建忠(Jian-Zhong Wu)

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(2014年)

https://doi.org/10.6841/NTUT.2014.00742

摘要

人們的興趣會因為環境的改變，而本身的興趣也跟著改變。我們的志趣常與工作上夥伴或共同學習的朋友具有較高的相似度。個人的喜好通常會受到周遭朋友的興趣影響而有所變化，而與舊朋友間的相似度，很有可能因為時間或距離因素而有所改變。為了結合使用者與好友間動態相似與情境資訊的應用，本論文提出結合使用者社群資料、氣候情境的考量與使用者親密度動態分析，稱為結合情境資訊與社群之動態時段協同過濾推薦系統(DTPCS)，篩選範圍內熱門且與使用者感興趣的景點來進行推薦，能減少相似度的計算。並考量在不同氣候情境下使用者需求的變化，探討氣候因素對喜好項目所帶來的影響。以及透過社群網絡的好友群，根據與朋友的打卡資訊來找出與使用者興趣喜好相似之朋友，給予不同的權重值來表示親密程度的不同，作為推薦的依據。最後計算出推薦序列，以Top-10之方式排序出推薦的清單。
根據模擬實驗之結果，本文所提出之DTPCS相較於DIRTD、TPPCF及TACF策略，可有效降低推薦誤差率與推薦計算時間。當行動用戶以平均每小時5 km/hr前進時，DTPCS之推薦誤差率平均優於其它策略約低35%。於推薦計算時間方面，DTPCS平均優於其它策略約少32%；而推薦覆蓋率方面，DTPCS平均優於其它策略約高27%。當以平均每小時50 km/hr前進時，DTPCS之推薦誤差率平均優於其它策略約低37%。於推薦計算時間方面，DTPCS平均優於其它策略約少36%；而推薦覆蓋率方面，DTPCS平均優於其它策略約高25%。依實驗結果顯示DTPCS在行動環境下能提供行動用戶相對穩定的推薦品質。

關鍵字

情境資訊 ； 社群 ； 協同過濾 ； 時間動態 ； 推薦系統

並列摘要

The people interest will often be changed by the dynamic of environment. We usually have high degree of similarity with partners of work or friends in the same school. Our preference usually be changed by around friend’s interests. The similarity of interest among old friends may change over time or distance. We proposed the Dynamic Time Periods Collaborative Filtering Recommendation System based on Contextual Information and Social Network (DTPCS), combining user community information, contextual information and user dynamic similarity. The system could query all popular Point of Interests (POIs) from the database recommend, it could reduce the computing of similarity. Moreover, we considered the changes of user requirements in different contexts, and explore the impact of contextual factors. Through the Community Network, the friends of similar interests could be found by the check-in information. The different contextual information will give different weight value to be as a basis for recommendation. Finally, the system sort the recommend sequence to be a Top-N recommendation list.
According to the simulation results, it showed that DTPCS we proposed had less recommendation error and system calculation time than DIRTD, TPPCF and TACF. When the mobile user’s velocity is 5 km/hr, in recommend error, DTPCS is less than the others about 35%. In recommend calculation time, DTPCS is less than the others about 32%. In recommend coverage, DTPCS is more than the others about 27%.The velocity is 50 km/hr, in recommend error, DTPCS is less than the others about 37%. In recommend calculation time, DTPCS is less than the others about 36%. In recommend coverage, DTPCS is more than the others about 25%.The experiment results showed that DTPCS was more suitable to be applied in the mobile environment.

並列關鍵字

Contextual Information ； Community ； Collaborative Filtering ； Temporal Dynamics ； Recommendation System.

參考文獻

[1] H. Zhang, L. Xu, H. Huang, and S. Gao, "Mining Spatial Association Rules from LBS Anonymity Dataset for Improving Utilization," in Proceedings of the 21st International Conference on Geoinformatics (GEOINFORMATICS), Kaifeng, China, 2013, pp. 1-6.

[2] J. Raper, G. Gartner, H. Karimi, and C. Rizos, "Applications of Location–Based Services: A Selected Review," Journal of Location Based Services, vol. 1, no. 2, 2007, pp. 89-111.

[4] J. B. Schafer, J. Konstan, and J. Riedl, "Recommender Systems in E-Commerce," in Proceedings of the 1st ACM conference on Electronic Commerce, Denver, Colorado, 1999, pp. 158-166.

[6] M. H. Kuo, L. C. Chen, and C. W. Liang, "Building and Evaluating A Location-Based Service Recommendation System with a Preference Adjustment Mechanism," Expert Systems with Applications, vol. 36, no. 2, 2009, pp. 3543-3554.

[8] Y. Zheng, L. Zhang, X. Xie, and W. Y. Ma, "Mining Interesting Locations and Travel Sequences from GPS Trajectories," in Proceedings of the 18th International Conference on World Wide Web, Madrid, Spain, 2009, pp. 791-800.