Bayesian Theory with Application to Customer Segment and Consuming Behavior Control for Online Business

貝氏理論應用於線上公司之顧客區隔與顧客監控

10.29977/JCIIE.200607.0004

羅淑娟(Shu-Chuan Lo)

層級貝氏 ； 有限混合模型 ； 馬可夫鏈蒙地卡羅法 ； 累積和管制圖 ； 顧客關係管理 ； Hierarchical Bayes ； finite mixture model ； Markov Chain Monte Carlo method ； CUSUM ； customer relationship management

工業工程學刊

23卷4期（2006 / 07 / 01）

303 - 310

英文

本文採用間隔購買時間(interpurchase time)模型分析顧客間的異質性，亦使用混合模型將顧客區隔成極活躍(super-active)、活躍(active)與不活躍(inactive)三種狀態。間隔購買時間模型與混合模型都是使用層級貝氏的概念，並透過馬可夫鏈蒙地卡羅法(Markov Chain Monte Carlo method)估計所需的參數。此外，本文採用累積和(CUSUM)管制圖以為顧客行為的監控工具，一般而言，管制圖通常是被企業用在產品品質的管制上，它提供一個合理的時間點去修正產品品質的偏差，本研究以活躍狀態的機率密度函數為基準，採用累積和管制圖的方式來監控顧客消費行為，圖中包含了個別顧客之整合後的消費間隔時間和最近消費時間（分析日與最後一次拜訪的時間間隔）訊息。最後，我們以一線上公司的實例實證，發現此模式之型Ⅰ誤差與型Ⅱ誤差分別小於5%與10%。

In this paper, we follow the model of interpurchase times to achieve heterogeneity across customers. We employ a mixture model to segment customers into three states: super-active, active and inactive. The interpurchase model and mixture model are solved by the hierarchical Bayes via Markov Chain Monte Carlo method. We employ CUSUM chart based on the density of active state to monitor consumer behavior. Factors on the CUSUM chart include interpurchase time and recency (the interval from the last visit day to the analyzing day) for individual customers. Control charts such as CUSUM chart have been used by industries to control product quality. They are useful time-series tools enable supervisors to correct the process in time when the chart shows a significant warning signal. In this research, we combined two information, interpurchase time and recency, from an individual customer and use CUSUM chart to control the consuming behavior. A real case study from an online company employed on CUSUM chart shows the type Ⅰ and type Ⅱ errors are less than 5% and 10%, respectively.

1. Allenby, G. M.,R. P. Leone,L. Jen(1999).A Dynamic Model of Purchase Timing With Application to Direct Marketing.J. of the American Statistical Association,94(446),365-373.
2. Barnard, G. A.(1959).Control Charts and Stochastic Process.J. of the Royal Statistical Society, Ser. B,21(2),239-257.
3. Carlin, B. P.,S. Chib(1995).Bayesian Model Choice via Markov Chain Monte Carlo Method.J. of the Statistical Society, Ser. B,57,473-484.
4. Evans, M.,I. Guttan,I. Olkin(1992).Numerical aspects in estimating the parameters of a mixture of normal distributions.J. Comput. Graph. Statist.,1,351-365.
5. Gelfand, A. E.,Smith, A. F. M.(1990).Sampling-Based Approach to Calculating Marginal Densities.Journal of the American Statistical Association,85,398-409.
6. Hughes, A. M.(2001).Strategic Database Marketing-the masterplan for starting and managing a profitable, customer-based marketing program
7. Juran, J. M.,A. B. Godfrey(1999).Juran's Quality Handbook
8. Montgomery, D.C.(1997).Introduction to Statistical Quality Control
9. Page, E. S.(1954).Continuous Inspection Schemes.Biometrika,41,100-115.
10. Shewhart, W.A.(1931).Economic control of quality of Manufactured Product
11. Theodossiou, P. T.(1993).Predicting shifts in the mean of a multivariate time series process: an application in predicting business failures.J. of the American Statistical Association,88(422),441-449.
12. Wang K. J.(2003).The Dynamic Predicting Model of Financial Distress for Real Estate Developers in Taiwan.