Impact of uncertainty in predicting the user's request on pushing

Pushing contents to users based on the predicted user preference cannot only improve user experience, but also boost network throughput by exploiting the excess resources. However, prediction is never perfect. Due to the uncertainties of predicting the user's request, the base station (BS) may waste resources on pushing unnecessary files before the arrival of user's request. In this paper, we investigate the impact of the uncertainty in predicting the content to be requested and the request arrival time on the average energy consumption of pushing. To this end, we first introduce a pushing policy with a priori known prediction uncertainty. Then, we derive the average energy consumption of pushing, and analyze the energy saving gain over traditional transmission method, where the BS serves a user right after the request arrives. Analytical and numerical results show that pushing with prediction uncertainty can save energy by optimizing the number of the pushing files, and the gain is remarkable for a user who has stronger preference among the predicted file list.