A Time Slot Count in Window Method Suitable for Long-Term Regularity-Based User Classification for Mobile Internet

Abstract

The mobile Internet proliferates and penetrates all aspects of daily life. It is crucial to codify methodologies to fit mobile-specific demands, e.g. regularity. Subscription-based mobile services require the analysis of regularity to capture and retain easy-come, easy-go users. The author proposes a method called the "time slot count in window" method (TCW-method) to efficiently capture regular users, while relaxing the explicit splits among time slots with visits. The author shows the case study result from 2001-2002 commercial service logs. A case study conducted in 2001-2002 with commercial news service subscribers shows that the relaxation in the proposed method loses 2-3 % in the recall rate of regularity, but still obtained 88-91 % recall rate. Considering the trade-off with the on-the-fly nature of coping with limited storage capacity, the author considers this method to be very promising for the mobile Internet.