A Mixture of Experts for Estimating Cellular Networks Failure Duration

Next-generation cellular networks herald a revolution in telecommunications infrastructure and services. 5G offerings include a wide range of services such as ultra-reliable, low-latency, scalable machine type and broadband communication. The technology promises to significantly impact industry, agriculture, healthcare, logistics, transportation, and several other sectors. Being a multi-service solution, 5G networks are highly complex. The presence of various hardware and software components generates multiple potential points of failure, leading to system interruptions and a degraded user experience. Thus, the ability to predict such failures and their duration can serve as a proactive network management mechanism, allowing counter measures to be planned ahead in order to mitigate such failures. In this context, this paper presents and evaluates a Mixture of Experts approach for detecting failures and estimating their duration in cellular networks, based on classification and regression models that assess Key Performance Indicators. The solution covers approximately 97% of the failure time in the evaluated scenario, demonstrating efficiency in detecting failures and estimating their duration.