一個依相似度做推薦的研究—以推薦中原夜市美食為例

A Study of Similarity-Based Recommendation using a Night Market as an Example

張祐翔(YU-HSIANG Chang)

指導教授 ： 夏延德

中原大學/工學院/資訊工程研究所/碩士(2014年)

https://doi.org/10.6840/cycu201400343

摘要

推薦機制起源於認知科學、逼近理論、資料檢索、預測理論。在1990年中期，推薦系統已經成為一個獨立的研究領域；除了學術中有大量研究外，在生活也與人們息息相關，比如日常在購物商場中店家會主動提供商品清單，或者是各式廣告跳出特價資訊提高消費者購買意願。
現在資訊爆炸出現許多大數據資料，導致基於歷史過往資料的推薦機制被大量使用，其中較為成功與廣泛地為協作推薦，協作推薦中可以區分成兩大塊，其一為基於使用者考量；其二為基於項目考量，但使用協作推薦必須克服三種問題主要為冷處理、稀疏問題、擴展性問題。
因此在本論文中我們為了克服冷處理問題，因此我們把基於歷史資料推薦方法改為使用相似性推薦方式來克服冷處理問題，最後並且根據相似性推薦機制製作一套應用在推薦中原美食系統。最後根據此系統我們分析使用者使用概況，並且根據已使用過系統使用者做面對面訪談，
針對相似性推薦方式成功實作在系統沒有使用者歷史資訊下作推薦，但是利用相似性推薦推薦機制來做中原美食推薦會有些問題，比方說店家屬性更動系統很難自動更新解決此問題、相同的使用偏好會得到相同推薦，此方式很容易造成使用者厭倦。

關鍵字

推薦機制、行動裝置、冷開始

並列摘要

Reference mechanism originates from cognitive theory, data index and forecasting theory. Reference system had become an independent field of research in mid 1990s. Other than its plenty academic research, reference mechanism has strong connection to human's life, such as commodity list provided by merchants in shopping malls, or sales information popping up in various ads targeting consumer's will of purchase.

With information explosion at modern times comes plenty of statistical data, resulting in the frequent use of reference mechanism based upon past information in the history, among which concordant reference is comparably successful and common. Concordant reference can be divided into two major parts: the one is tailored for users, the other for items, whereas it is necessary to overcome three problems upon its use, mainly cold approach, sparse and extensive problems.

Therefore, with view to surmounting cold approach problem in this thesis, we revised reference methods based upon historic data to the ones that invoke similarities for grappling with cold approach problems. Finally, we made a system applied to reference to central cuisine through similarity reference mechanism. In addition to that, we analyzed user's utilizing conditions and interviewed with those who have already navigated the system.

As far as the success of similarity reference methods without the system being used along with historic data is concerned, there are some problems of using similarity reference mechanism to conduct central cuisine reference, though. For instance, business attributes update system has difficulty automatically handling the problems, and the identical use preferences lead to identical references, which upsets users.

並列關鍵字

recommender system ； mobile system ； cold-start

參考文獻

Fabian Abel, Ig Ibert Bittencourt, Evandro Costa, Nicola Henze, Daniel Krause, and Julita Vassileva, “Recommendations in Online Discussion Forums for E-Learning Systems”, Learning Technologies, IEEE Transactions on (Volume:3 , Issue: 2 ), pp. 165-176, April-June 2010.

Sanjay Ghemawat, Howard Gobioff, and Shun-Tak Leung. “The Google File System,” 19th ACM Symposium on Operating Systems Principles, Lake George, NY, October, 2003.

Kazunari Sugiyama, Kenji Hatano, and Masatoshi Yoshikawa. “Adaptive Web Search Based on User Profile Constructed without Any Effort from Users,” WWW 2004, New York, USA, pp. 675-684, 2004.

C. Cornelis, J. Lu, X.. Guo, and G. Zhang. “One-and-only item recommendation with fuzzy logic techniques,” Information Science, vol. 177, pp. 4906-4921, 2007.

C. Porcel, A. G. López-Herrera, and E. Herrera-Viedma. “A recommender system for research resources based on fuzzy linguistic modeling,” Expert Systems with Applications, vol. 36, no. 3, pp. 5173- 5183, 2009.