Customer Service Escalation Early Warning System: A Subsampling Approach

• Activity Number: 312
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Marketing
• Author(s): Jiabin Zhao*+
• Companies: Cisco Systems
• Keywords: data sampling ; logistic regression ; clustering ; early warning system ; networking ; professional service

Abstract

The networking service, both advanced and technical, plays an important role in the growth of the industry. However, with the massive amount of systems and solutions deployed, the cost of the service, especially to support the customer escalation, has skyrocketed. It becomes critical to understand customers and networks to support decision-making processes at each stage of the service. We have developed a statistical framework for the assessment of the likelihood of the customer escalation for the services. The framework contains the data engineering, clustering, sampling, and logistic regression analysis for the big data generated from the networks and service process. This paper mainly focuses on the sampling methodology and algorithm. Re-sampling technique is commonly used for the class imbalanced data. It provides many advantages but isn't efficient and practical. We present that combining different expressions of the re-sampling approaches is effective and practical to real world problems. We have applied the methodology combining with the framework to the class imbalanced data of Cisco to identify the underlying factors to cause the escalation. The results are quite accurate.