網際網路服務業客戶流失預測模式之研究

網際網路服務市場競爭激烈，客戶流失日益嚴重。過去對客戶流失之研究，大多從客戶滿意度或客戶忠誠度等單一角度進行探討，缺乏整合性之研究。本研究利用資料探勘之決策樹分類技術，結合客戶特性、客戶滿意度、客戶忠誠度、及服務品質等4個構念，共14項變數，建立客戶流失預測模式。研究之樣本，為某網際網路服務公司的ADSL企業用戶，有效資料12,177筆，其中87.5%作為訓練資料，共10,655筆，其餘的12.5%，共1,522筆，作為測試資料。研究結果顯示，客戶類別、客戶忠誠度、客戶滿意度、服務品質等均會影響客戶是否流失，其中客戶忠誠度對於預測網際網路服務業客戶是否流失效果最顯著，而本研究之數值型變數有較佳的預測力，配合適當的探勘策略，可以提升預測準確度，本研究預測準確度達95.99%，優於一般相關研究。本研究最後獲得影響客戶流失的9個主要變數，分別為已租用時間、客服服務次數、頻寬類型、折扣比率、申請時間、購買金額、斷線次數、用戶類別、最近購買日等，可作為客戶流失預測模式的輸入變數。

關鍵字

網際網路服務業；客戶流失預測模式；資料探勘；決策樹

並列摘要

The competition of internet service market is vigorous. The customer churn is serious. The past related researches always focus on a single view, such as customer's satisfaction or the customer loyalty, mostly lack of integrating research. This research used data mining techniques to build ISP customer churn prediction model with the information of customer characteristics, customer satisfaction, customer loyalty, and service quality of ISP customers. This model can provide information for enterprise to adjust their strategies to retain customers, increase the customer royalty, decrease customer churn rate, and gain more profit. The sample of this research is the ADSL customers of a leading telecom company in Taiwan. The total valid data is 12,177 records, the 87.5% as the training data, and the other 12.5% as the testing data. We found that the customer royalty is the most sensitive factor for the churn prediction model, and the numerical data has higher hit rate than category data. The correct data mining strategy can promote the hit rate. The hit rate of our prediction model is 95.99%, better than the related model. The most significant factors of this model are ”The leased time”, ”Customer service records”, ”Usage rate”, ”Discount rate”, ”Applying time”, ”Monetary amount”, ”Disconnection times”, ”Customer type”, and ”Recently purchased date”.

並列關鍵字

Information Service Provider ； Customer Churn Prediction Model ； Data Mining ； Decision Tree

參考文獻

台閩地區電腦應用概況報告

Google Scholar

Berry, Michael J. A.,Gordon Linoff(2001).Mastering Data Mining, The Art & Science of Customer Relationship Management.John Wiley & Sons, Inc.

Google Scholar

Bolton, Ruth N.(1998).A Dynamic Model of the Duration of the Customer`s Relationship with A Continuous Service Provider: The Role of Satisfaction.Marketing Science.17(1),45-65.

Google Scholar

Cai, Y.,Deilami, I.,Train, K.(1998).Customer retention in a competitive power market: Analysis of a `double-bounded plus follow-ups` questionnaire.Energy Journal.19,191-215.