Cost-Sensitive-Data Preprocessing for Mining Customer Relationship Management Databases.

The article discusses the computer industry's efforts to reduce the number of customers who switch service providers. Research, in the cellular telephone industry for example, indicates that 30% of the customers switch to another provider. The cost to get a lost customer back is much higher than the cost of retaining a customer. Computer systems are being developed that will identify a customer that is likely to churn, or switch, and formulate plans to fight the problem. Classifying customers as likely to switch is cost-sensitive and data from such customers contains a small portion of positive data and a large portion of negative data.