This paper presents the potential for non-invasive and inexpensive monitoring of communication and mobility network conditions using Norwegian mobile network usage data. We transform network data into models within a reservoir computing framework and then measure the model's performance on proxy tasks. Experiments demonstrate how the performance of these proxy tasks correlates with network conditions. A key advantage of this approach is its use of readily available data sets and the availability of a reservoir computing framework, which is inexpensive and adaptable to most algorithms. Anonymized and aggregated mobile network usage data is available in multiple daily snapshots. This data can be processed by weighted networks, and reservoir computing allows the use of untrained weighted networks as machine learning tools. Networks initialized with echo state networks (ESNs) project input signals into a high-dimensional space, with a single trained layer. This approach consumes less energy than deep neural networks that train all network weights. We train ESN models using neuroscience-inspired tasks and demonstrate that performance varies depending on the specific network configuration and degrades significantly when the network is perturbed. While serving as a proof of concept, we anticipate that it could also be used for real-time monitoring and identifying vulnerabilities in mobile communication networks and transportation networks.
