應用時間加權於可重複序列之研究-以預測線上顧客消費狀態為例

在以一對一滿足顧客需求為競爭關鍵的時代中，企業管理者若能有效掌握顧客的消費行為，適時提供正確的服務來滿足顧客的需求，便能提升企業競爭優勢，也因此瞭解顧客消費行為的變化就成為企業獲利與否的關鍵因素。而近年來序列樣式探勘為探討顧客行為規則的研究焦點之一，但是鮮少於討論不同的時間點影響不同程度的行為變化之議題。有鑑於此，本研究將建構一預測顧客行為狀態的模式，此模式結合了時間加權概念與可重複序列樣式探勘，其中，時間加權概念主要目的是為了解決顧客行為狀態會隨著時間週遭因素的影響，而使得序列資料的變異性較大，而導致分析結果表現較不為理想。而可重複序列則可以單獨對一顧客序列作探勘，這樣可以針對高價值或高忠誠度顧客進行一對一的分析，以提高顧客滿意度。本研究以國內某線上音樂公司為例，預測顧客其次時點的消費狀態。在時間加權方面導入了越近期資料給予越大權重和同屬性期間加權，管理者便可依照時間不同的重要程度將顧客序列予以分割，並且給予不同的權重大小，以產生新的且符合管理者所需的預測規則，線上公司可依此預測規則針對顧客做預先個人化服務配置，對線上公司在提升顧客服務行動上，提出一具體可行方案。

關鍵字

二元序列樣式探勘；時間加權模式；可重複序列

並列摘要

In this period when we regard the satisfaction of one to one customer’ needs as a key to competition, we can raise the advantage for business competition if the business managers can effectively control the consuming behavior of customers and offer right services to satisfy customers’ needs at an appropriate time. Therefore, it becomes the essential factor in the profit of enterprises that we understand the change of consuming behavior of customers. In recent years, sequential pattern mining has been one of the research focuses on the rule of customer’s behavior. Nevertheless, it has been hardly discussed that different timing could affect the change of behavior to some extent. Because of that, this research will construct a model of predicting the state of customer’s behavior, which combines the time-weighted concept and re-counting sequence mining. In this model, the main purpose of time-weighted concept is to resolve the problem that the status of customer’s behavior varies in time, which results in obvious variety of sequence data to make analysis performances less desirable. And the re-counting sequence can mine for single customer-sequence, which is able to analyze those high-value or highly loyal customers in a one to one way and raise customer satisfaction. This method divides customer-sequence into several partitions according to time and gives each partition different weight to produce new prediction rules that interest managers.

並列關鍵字

Sequential Patterns Mining ； Time-weighted ； Re-counting Sequence

參考文獻

[4] 陳仕昇、陳彥良、許秉瑜 ,”以可重複序列探勘網路瀏覽規則之研究,”資管評論,第九期 1999。

[7] 龔書賢，應用基因演算法及權重項目法於關聯法則挖掘之研究，碩士論文，元智大學工業工程與管理學系，桃園，2003。

[9] Agrawal, R. and Srikant, R. ,“Fast Algorithm for Mining Association Rules ,” in Proc.20th VLDB conference ,Santiago, Chile, 1994.

[14] Cooley, R., Mobasher, B., Srivastava, J. , “Data preparation for mining world wide web browsing patterns,” Journal of Knowledge and Information Systems ,vol.1 (1) 1999,pp.5-32.

[21] Pei, J. , Han, J., Pinto, H., Chen, Q., Dayal, U. and Hsu, M.-C., “PrefixSpan:mining sequential patterns efficiently by prefix-projected pattern growth,” in Proc. of the 17th International Conference on Data Engineering, 2001, pp. 215-224.

國際替代計量

應用時間加權於可重複序列之研究-以預測線上顧客消費狀態為例

未授權

主題瀏覽