Performance and Usage Patterns of Mobile Networks

The popularity of mobile devices and the availability of various services over mobile cellular networks has increased over the past twenty years. Over time, mobile cellular network technologies have evolved, and the performance of wireless links from mobile devices to the core networks is increasing. Mobile applications and services require differ- ent network qualities to meet users’ expectations and increase the Quality of Experience (QoE). To support the increased number of users, and to deliver the capacity required by applications, mobile networks have become complex systems. The demand for high-quality experience in mobile cellular networks is in the interest of both end-users and providers. However, mobile network performance is affected by a multitude of network features. This includes radio technology, network bandwidth and coverage, signal strength, mobility, throughput, latency, and data usage patterns of users. This thesis analyzes mobile cellular network performance and usage patterns. We apply different data analysis methods and use various datasets collected through crowdsourcing and testbeds. We study various features of mobile networks and their effect on mobile network performance. We propose an estimation method for QoE in web browsing and discuss factors affecting web-flows performance in mobile networks. We present different models based on machine learning that predict network throughput, cluster, and classify mobile users’ data usage patterns. This thesis contributes to the evolving mobile networks by studying various network features that determine the performance of mobile networks and the data usage patterns of mobile users. The large-scale crowdsourced mobile network measurement datasets provide valuable input for understanding factors affecting the performance and quality of mobile networks. The study on the data usage patterns of mobile users provides significant input for understanding mobile users’ data usage patterns and behavior across different countries. The classification model on network stability and data usage patterns can be valuable input for network resource optimization. The study conducted on the feasibility of teleoperated driving and correlation-based network feature mapping shows how crowd-sourced datasets can be used to analyze different uses cases in mobile networks