A Binary Time Series Model of LTE Scheduling for Machine Learning Prediction

Abstract

In today's Third-Generation Partnership Project (3GPP) Long-Term Evolution Advanced (LTE-A) cellular radio networks, battery lifetime is critical for mobile devices. During time intervals of no user data transmit or receive activity, energy for receiving and processing irrelevant control information in a mobile device could be saved. Therefore, we propose a binary time series model at 1 ms transmission time interval (TTI) granularity to predict the control channel information. To assess the predictability of the proposed time series, we apply three well-known machine learning (ML) algorithms combined with a non-intrusive cost-sensitive classification (CSC) scheme. Predictions of the proposed time series model successfully reach false negative rates (FNRs) below 2%.