Machine learning-based quality evaluation of 5G wireless network in Smart Grid

Effective network quality level evaluation can provide a reliable reference for the link selection of business transmission. The network quality level evaluation based on specific businesses can provide more accurate network quality level references for different services. To this end, this paper proposes a machine learning-based network quality evaluation model. The model first discriminates the type of business by a random forest algorithm based on the network needs of the business. Then, the temporal properties and nonlinearity of the evaluation index data were fitted by LSTM, using multi-head attention (MA) to learn and obtain the evaluation index weights under different businesses, so that the network quality rating evaluation results under the corresponding business are obtained. In this paper, the network quality level evaluation under the media services and precision control services in the smart grid is discussed, and the proposed model is compared with the Support Vector Machine model (SVM) and Recurrent Neural Network-Multi-head Attention (RNN-MA) model. The experimental results show that the proposed LSTM-MA model has the best prediction accuracy.