Hierarchical user interest model based on large log data of mobile internet

The development of mobile internet brings a new perspective for user interest model. Mobile logs record more behaviors which can build a more comprehensive model for personalized service or recommendation system. Comparing the VSM and semantic network, we choose Concept Hierarchy to model user interest considering hierarchical granularity relations and make the user interest vivid and clear. In the background of large scale data, we analyze the characteristic of mobile internet and propose a mechanism based on session, so as to make the process of modeling linear and superimposed. In addition, we employ unsupervised learning method to avoid storing large historic data. At the end, we import time decay function and spreading activation model to trace the change of user interest. Experiential results confirm the feasibility and availability.