• Journal of Intelligence and Information Systems
• /
• v.10 no.3
• /
• pp.115-132
• /
• 2004

Recently, as most companies recognize the importance of the customer relationship management, they strongly believe that they must know who their customers are. The job of a customer is very important information for us to understand the customer. However, since most customers are reluctant to reveal them-selves, they do not let us know their jobs, and even provide false information about their jobs. The target domain of our research is mobile telecommunication. In this research, we developed a system that identifies the customer's job by utilizing the Call Detail Record. Using artificial neural networks, we developed a two-step Job Identification System. In the first step, it identifies the four job classes, then in the second step, it subdivides these four job classes into seven jobs. The accuracy of identifying the seven jobs was $71.9\%$.

• Proceedings of the Korea Inteligent Information System Society Conference
• /
• 2001.06a
• /
• pp.243-252
• /
• 2001

Recently, as most companies recognize the importance of the customer relationship management, they strongly believe that they must know who their customers are. The job of a customer is very important information for us to understand the customer. However, since most customers are reluctant to reveal themselves, they do not let us know their jobs, and even provide false information about their jobs. The target domain of our research is mobile telecommunication. In this research, we developed a system that identifies the customer's job by utilizing the Call Detailed Record. From the Call Detailed Record, we extracted such variables as 'Average Monthly Payment'and 'Age of the Customer'and so forth. Moreover, we generated many summary variables and derived variables such as 'Number of Calls during Work Hours in Weekday', and 'Ratio of Calls from other Mobile Telephone'. Using artificial neural networks, we developed a two-step Job Identification System. In the first step, it identifies the four job classes then in the second step, it subdivides these four job classes into seven jobs. The accuracy of identifying the seven jobs was 69.1%.