Predicting and Analyzing IoT Performance Based on Improved Resnet

With the increasing maturity of smart communication technologies and the rapid popularization of mobile IoT devices, technological development has gradually stepped into the world of the Internet of Everything, and the mobile Internet of Things (IoT) is booming, resulting in numerous applications of great significance. The complex and changing environment of mobile communication requires that mobile IoT applications can en-sure reliable delivery, more agile and accurate performance pre-diction methods are required. Outage probability (OP) is an important metric if one wants to evaluate the performance of mobile IoT networks. In this paper, we investigate the OP performance of mobile IoT communication networks based on an improved deep residual network (Resnet) and propose an prediction algorithm for OP based on the improved net-work. We first study the mobile OP performance and derive the accurate OP expression under a specific channel. Then, an improved residual network is designed to improve the prediction accuracy, the pooling layer of network is removed due to the characteristics of mobile communication data. The pro-posed prediction method based on improved residual network is compared with traditional convolutional neural network (CNN), generalized regression (GR) and radial basis function (RBF) methods. The simulation results verify that the proposed improved residual network prediction method can obtain better prediction results than the above traditional methods. The prediction accuracy of the Resnet method is improved by 44%.