運用隱性回饋趨勢推薦產品方法之研究

An interval-based Recommender System with Implicit Feedback

趙思棋(Szu-Chi Chao)

指導教授 ： 許秉瑜

國立中央大學/管理學院/企業管理學系/碩士(2011年)

摘要

在現在電子商務蓬勃發展的時代中，消費者可以透過網路搜尋到許多產品資訊，但卻也讓他們面臨到資訊超載的問題。企業為了能節省消費者搜尋產品的時間、提高消費者滿意度並掌握其消費動向，經常會仰賴推薦系統以加強與消費者之間的關係。透過推薦系統可以分析消費者的消費習慣以及產品偏好，再依此提供合適的商品給消費者，以期待能提高消費者的購買意願。但傳統的推薦系統大多屬於靜態的，一般會將消費者某一期間的歷史資料進行分析並找出其喜好，但卻未考慮到消費者的喜好並不是持久不變，而是具有時間性。
　　為了改善傳統推薦系統的推薦品質，本研究將考慮時間因素，提出了以時間為基礎的內容導向式推薦系統架構。本研究的目的在於觀測消費者隨著時間變化的交易行為並找出消費者的興趣趨勢，進而達到推薦符合消費者需求的產品。透過建置一套機制將消費者的交易行為轉成偏好的程度，再結合時間因子對分數進行調整，以分析消費者偏好的趨勢，有效的進行個人化推薦。本研究以某手機服務網站的交易資料做為實證資料以驗證其推薦效果，其結果顯示本方法較未考慮時間因素的傳統方法其精準度高出83.5%，顯示本研究方法更能有效的預測消費者在未來期間可能會購買的產品類別。

關鍵字

內容導向過濾法 ； 隱性回饋 ； 時間基礎推薦系統

並列摘要

The e-commerce is booming through rapid expansion of Internet. However, consumers can find many products and related information from the Internet, but also they face a problem to choose their right Internet within short time. Enterprises usually rely on a recommender system to strengthen the relationship with consumers, such as saving consumers’ search time, improving customer satisfaction and tracking the consumer behavior. Through recommender system enterprises can analyze the consumer behavior and product preferences and can provide the right goods to consumers to enhance the consumer''s purchase intention. However, most of the traditional recommender systems are static which identify consumers’ preferences by analyzing their historical data of the certain time period and those methods don’t consider a time factor that consumers’ preferences might change over time.
This study propose a time-based content based approach for recommendation that considers time factors to the operation of the recommender system to improve the efficiency of the traditional recommender systems. The purpose of this study is to recommend the products that meet the consumers’ demand, by observing the changes of consumers’ transactions over time and identifying trends in consumers’ interests. We establish a model to convert the consumer transactions to the degree of preference and combine with the time factor. By using this model, we analyze trends in the consumer purchase interests and provide personal recommendation efficiently. This study uses the mobile service transaction database to assess the performance of the proposed approach. The experiment result shows that the proposed approach can predict the consumers’ purchase interest efficiently in the products’ category level.

並列關鍵字

content-based recommender system ； time-based recommender system ； implicit feedback

參考文獻

林建煌，消費者行為二版，華泰文化事業股份有限公司，2007。

Adomavicius, G.., and A. Tuzhilin, Toward the next　generation of recommender systems: a survey of the state-of-the-art and possible extensions. IEEE Transactions on Knowledge and Data Engineering, 17(6), 734-749, 2005.

Andrews, J. C., S. Durvasula and S. M. Akhter, A Framework for Conceptualizing and Measuring the Involvement Construct in Advertising Research, Journal of Advertising, 19(4), 29, 1990.

Balabanovic,M., and Y. Shoham, Fab: Content-based, Collaborative Recommendation. Communications of the Association for Computing Machinery, 40(3), 66-72, 1997.

Cheung, K.W., K.C Tsui, and J. Liu, Extended latent class models for collaborative recommendation. IEEE Transactions on Systems, Man and Cybernetics, Part A: Systems and Humans, 34(1), 143-148, 2004.