A Space-Optimal Month-Scale Regularity Mining Method with One-Path and Distributed Server Constraints for Mobile Internet

Abstract

Mobile Internet becomes a first-class citizen of Internet in many advanced countries. As increased penetration leverages mobile application business opportunities, it is important to identify methodologies to serve mobile-specific demands. Regularity is one of the important measures to retain and enclose {\it easy-come, easy-go} mobile users. It is known that a user with multiple visits in one day with a long interval has a larger revisiting possibility in the following month than the others. The author investigates the minimum number of bits to incorporate this empirical law in order to cope with the two major mobile restrictions: distributed server environments and large data stream. The author shows that the method with 2+1 bits can provide usable results to classify regular users in the case study. It gives the lower-bound of memory needed to identify revisiting users under mobile-specific constraints.