顧客流失分析模式之個案研究－以臺灣H健康休閒俱樂部為例

A Case Study of Customer Churn Mod-An Example of H Health Club in Taiwan

陳麒文(Chi-Wen Chen)

《臺灣體育運動管理學報》 5期 (2007/09) Pp. 124-149

https://doi.org/10.6547/tassm.2007.0007

摘要

本研究之主要目的包括瞭解臺灣某家健康休閒俱樂部（以下均以H club稱之）之顧客組成結構、利用資料採礦分類技術（鑑別分析、羅吉斯迴歸、人工類神經網路、多元適應性雲形迴歸）來建立H club之顧客流失分析模式、與經由顧客流失分析模式來瞭解流失顧客之重要特徵。本研究利用H club所提供的資料來進行實證研究，共計1,287筆，在剔除資料不全與資料內容不合理的資料後，共有1,152筆。本研究之結果如下：
一、H club之顧客組成結構以女性會員(61.72%)佔大多數、平均年齡為38.16歲且多集中在25歲~45歲(69.03%)、平均會齡為2.63年且多集中在二年以下(52.86%)、入會金額以0元為最多(50.52%)、月費金額以2,500元為最多(36.37%)、居住地區則以臺北市居多(81.68%)、付款方式則多為現金(59.46%)。
二、本研究所提出之顧客流失分析模式之建構程序，主要的目的是希望透過四種不同分類方法的比較，來求得一個最佳的區別模式。此外，為驗證所提模式之有效性，本研究利用H club所提供的資料來進行實證研究，結果顯示，多元適應性雲形迴歸的整體分類績效為86.52%，具有極佳之分類效果，是一項
值得建議使用的工具。
三、總體而言，以整體正確判別率為最高之多元適應性雲形迴歸所建構之顧客流失分析模式，其流失顧客之特徵為年齡介於30~35歲、會齡為一年以下、月繳2, 500元的月費、且以現金支付費用的會員。

關鍵字

資料採礦 ； 顧客流失 ； 鑑別分析 ； 羅吉斯迴歸 ； 人工類神經網路 ； 多元適應性雲形迴歸

並列摘要

The main purpose of this study was to understand the customer structure in a certain Taiwanese health club (the following substituted as H club) and to establish its customer churn model through data mining classification technology (discriminant analysis, logistic regression, artificial neural networks, multivariate adaptive regression splines) as well as to realize the significant characteristics of churn customer through the customer churn model. The data used was provided by H club to perform the empirical research, and the original data were 1,287 records. After eliminating the incomplete data and the unreasonable data, there were 1, 152 records totally. The results were as followed:
1. In the customer structure of H club, the majority of gender was female member taken 61.72 percent of total. While the average age was 38.16-year-old, the range of age was from 25-year-old to 45-year-old taken 69.03 percent. The duration time of membership was 2.63-year in average and most centralized in two-year-below (52.86%). The initiation fee was 0 New Taiwanese dollar taken the major part of total (50.52%). Otherwise, the monthly fee was 2,500 New Taiwanese dollars in majority (36.37%). With the residential area, the most members were living in Taipei city (81.68%). Finally, the method of payment was in cash (59.46%).
2. The constructional process of churn model in this study was by way of four classification methods to obtain the one best discriminating mode. Besides, in order to verify the effectivity of the discriminating mode, this study used the data provided by H club to perform the empirical research. The result found that wholecorrect classification rate was 86.52% in the churn model using the multivariate adaptive regression splines, which had the best utility and been suggested as a worth tool to use.
3. By all accounts, the characteristics of churn customer through the customer churn model using the best way of multivariate adaptive regression were the members of the age between 30-year-old and 35-year-old, the duration time of membership of one-year-below, the monthly fee of 2,500 New Taiwanese dollars, and the method of payment of cash.

並列關鍵字

Data Mining ； Customer Churn ； Discriminant Analysis ； Logistic Regression ； Artificial Neural Networks ； Multivariate Adaptive Regression Splines

參考文獻

Berry, M. J. A.,Linoff, G. S.(1997).Data mining techniques: For marketing, sales, and customer support.New York:Wiley Computer.

Berson, A.,Smith, S.,Thearling, K.(2000).Building data mining applications for CRM.New York:McGraw-Hill.

Browne, M. J.,Hoyt, R. E.(1995).Economic and market predictors of insolvencies in the property-liability insurance industry.Journal of Risk and Insurance.62,309-327.

Cabena, P.,Hadjinaian, P.,Stadler, R.,Verhees, J.,Zanasi, A.(1997).Discovering data mining from concept to implementation.New Jersey:Prentice Hall PTR.

Cios, K.,Pedrycz W.,Swiniarski R.(1998).Data mining methods for knowledge discovery.Boston:Kluwer Academic.