Fault Prediction Method for 5G Bearer Network Based on Graph Random Walk

With the development of 5G networks, the number and complexity of network devices have increased dramatically. Locating and predicting faulty network devices is an important way to ensure the normal operation of 5G networks. Existing methods usually require the use of large amounts of real-time data, which is unacceptable for widely distributed network devices. We propose a 5G bearer network fault prediction method based on graph random walk. By analyzing the association relationship of network devices’ alarm information, we construct an alarm association graph. Use the graph random walk method to learn alarm rules from historical alarm information and use it to predict faulty devices. We verified our method on real data from network service providers and achieved better results than existing methods.