考量近度、頻度與值度的序列樣式分析方法

A Method to Discover Sequential Patterns with Considering Recency, Frequency and Monetary

吳欣怡(Shin-Yi Wu) ； 陳彥良(Yen-Liang Chen)

《電腦與通訊》 145期 (2012/06) Pp. 94-101

https://doi.org/10.29917/CCLTJ.201206.0014

摘要

序列樣式探勘在電子商務領域是一項有用的技術，在電子書城平台亦然。本文針對電子零售業之需求，將行銷學的RFM觀念融入序列樣式探勘技術，可針對使用者點選記錄、閱讀記錄、購買記錄進行分析，適當設定近度、頻度、值度之條件範圍，所獲得之RFM序列樣式可應用於改良電子書城平台架構設計、個人化書籍推薦、個人化電子型錄設計等。本文提出一個RFM序列樣式探勘演算法，以真實零售業交易記錄驗證其成效，並提出一個RFM序列樣式探勘之應用方針，比對不同區隔之序列樣式可進一步了解資料的隱含意義，提供更有價值的決策資訊。

關鍵字

資料探勘 ； 序列樣式 ； 限制型探勘 ； 近度頻度值度區隔

並列摘要

Sequential pattern mining is useful in electronic commerce. It is also helpful to e-book store. To fulfill the requirement of e-retailing, this paper incorporated the RFM concept from marketing into sequential pattern mining. With adequately setting the selection criteria of recency, frequency, and monetary, the technique can analyze the user click stream, reading log, and also purchasing log. The discovered RFM sequential patterns can be applied into improving the design of Web site, personalized book recommendation, personalized e-catalog design, and so on. This paper proposed an algorithm to discover RFM sequential patterns. We use the real dataset from a retail chain to evaluate the proposed algorithm, and then propose an application guideline for RFM sequential pattern mining. Comparing the pattern groups of different segments help understand the implicit information in datasets, and then achieves the purpose of decision support.

並列關鍵字

Data Mining ； Sequential Pattern ； Constraint-based Mining ； RFM Segmentation