Joint Statistical and Machine Learning Approach for Practical Data-Driven Assessment of User Throughput Quality in Microcellular Radio Networks

Abstract

The provision of enhanced user experience with higher-data transmission speed over the radio frequency channels was one of the core objectives of evolving Long Term Evolution (LTE), which is a 4th generation of cellular broadband communication standard. LTE was developed to supports data transmission speed up to 300 Mbps and 75 Mbps in downlink and uplink, respectively, utilizing a robust multiplexing technique. For mobile radio networks operators to maintain high operability around the ever increasing and demanding subscribers, periodic evaluation, quantitative estimation and analysis of network performance is pivotal. There exist a quantum of previous research works on LTE system network performance that have been conducted both at spatial domain and temporal domain, but the authors concentrated their studies either on practical LTE radio coverage issues only or on general LTE network performance issues using analytical/simulation techniques. In this contribution, a combined statistical and machine learning approach is proposed and applied to provide an in-depth user data throughput performance of operational LTE networks in relation to signal coverage and signal quality parameters, using in typical microcellular built-up terrains as case study. Specifically, the impact of transmitter- receiver communication distance, received signal reference power, received Signal reference quality, Signal to Noise Ratio, Received Signal Strength and Channel Quality Indicator on user throughput over Radio Link Control, Physical Downlink Shared Channel and Packet Data Convergence Protocol layers have been shown using professional TEMS investigation tools. Particularly, for the purpose of effective LTE system network planning and management, the most influencing signal coverage and signal quality parameters on user throughput performance are also examined and shown in the framework of Self-organsing map (SOM) and statistical correlation stratagems.