發展一個序列樣式變化之偵測模型-考慮間隔時間因素

Detecting the change of customer behavior based on Time-Interval Sequential Patterns

林昭妏(Chao-Wen Lin)

指導教授 ： 蔡介元

元智大學/工程學院/工業工程與管理學系/碩士(2007年)

https://doi.org/10.6838/YZU.2007.00243

摘要

在現今快速變化的消費市場中，管理者若能有效掌握顧客的消費行為模式便能適時制定有效的行銷策略並且提供正確的服務以滿足顧客的需求。因此，了解顧客的消費行為便能增加企業的獲利並提昇其競爭能力。目前已有不少研究在探討顧客行為的規則，也有應用序列樣式探勘技術探討在兩個不同時間區間之下的消費行為的變化，然而卻很少有研究針對序列樣式中項目間的間隔時間做探討。因此，本研究提出TI-AprioriAll演算法用以挖掘包含間隔時間的序列樣式，並且應用在兩個不同時間區間的消費行為變化模式做探討。挖掘出兩個不同時間區間的間隔時間序列樣式後，兩兩比對並且計算其改變的程度，再將所有間隔時間序列樣式歸納為三種改變的類型(即emerging sequential patterns, unexpected sequence changed and added/perished sequential patterns)。最後，篩選出顯著的消費改變類型做分析討論，以供決策者參考制定行銷策略。

關鍵字

資料探勘 ； 序列樣式探勘 ； 變化探勘

並列摘要

Current business undergoes the challenge of a rapid evolving market in which customer’s needs are changing over time. There have been several studies in data mining field focusing on mining changes of sequence in different time-period databases. Analyzing these changed sequences provide useful information for decision makers to make better marketing strategy to attract more customers. However, the time-intervals between successive itemsets in the sequences are not considered in their proposed methods. In fact, changed patterns with time-interval can reveal important time information for decision makers to take the right actions at the right time. Therefore, this research develops a time-interval sequential pattern change detection method to derive the change trend of customers’ behavior in two periods. First, the TI-AprioriAll algorithm, which integrated Apriori-like algorithm and k-means cluster algorithm concepts, is proposed to generate time-interval sequential patterns from two different time-period databases. Then, a time-interval sequential pattern is clarified as one of three change types (emerging sequential patterns, unexpected sequence changed and added/perished sequential patterns) through the proposed sequential pattern matching method. Finally, a set of time-interval sequential patterns with significant change are retrieved for further business analysis.

並列關鍵字

Data Mining ； Time-Interval Sequential Pattern Mining ； Change Mining

參考文獻

謝玉貞，「應用序列樣式探勘技術於行為變化之研究」，碩士論文，元智大學工業工程與管理研究所，民國94年。

林柏伸，「行動環境下之使用者行為樣式研究—以二維度序列型樣進行探勘」，碩士論文，中原大學資訊管理研究所，民國92年。

Altiparmak, F., Ferhatosmanoglu, H., Erdal, S. and Trost, D. C., “Information Mining Over Heterogeneous and High-Dimensional Time-Series Data in Clinical Trials Databases,” IEEE Transactions on Information Technology in Biomedicine, Vol. 10, pp. 254-263, 2006.

Chen, M.C., Chiu, A.L., Chang, H.H., “Mining changes in customer behavior in retail marketing,” Expert Systems with Applications, Vol. 28, pp. 773-781, 2005.

Chen, Y. L., Chiang, M.C., Ko, M.T., “Discovering time-interval sequential patterns in sequence databases ,” Expert Systems with Applications, Vol. 25, pp. 343-354, 2003.