Data-driven active session identification for LTE user-perceived QoS analysis

Abstract

While the telecommunications landscape is undergoing a significant transformation with the advent of 5G technology, the continued importance of Long Term Evolution (LTE) is also emphasized due to its widespread adoption and reliability. In this circumstance, mobile network operators must continue to uphold their obligation to ensure Quality of Service (QoS) for LTE users. The actual LTE Base Station (BS) signal measurement data can be effectively exploited in the evaluation of user-perceived QoS. In this work, we utilize Downlink Control Information (DCI) data obtained from LTE BSs through a recently developed platform. The DCI data contains downlink information experienced by all users within the cell, but to assess user-perceived performance, it is necessary to distinguish each active session. We call the grouping of DCI messages that correspond to a continuous service in one active session ‘bundling’. While previous bundling methods have mostly focused on the time gap between DCI messages, we extract features based on the LTE standard that DCI can exhibit at the start of an active session. By combining these features with a probabilistic approach, we establish criteria for implementing bundling. In addition, through the proposed bundling methodology, we analyze the active session duration, number of active sessions, cell edge user performance, etc. The results validate the effectiveness of our bundling approach.