Accurately predicting the traffic status of devices in the network is of great significance for network measurement and management, which is helpful to further improve the quality of network service and user experience. However, due to the strong nonlinearity and uncertainty of network traffic, it is still difficult for traditional statistical methods or machine learning methods to obtain better prediction results. In order to further improve the accuracy of network traffic prediction, a local context information enhanced attention mechanism is designed and implemented. The input sequence is converted into Q and K through convolution calculation. So, the time series can be interpreted from a microscopic perspective, which improves the local awareness of the attention mechanism. Furthermore, the proposed attention mechanism is combined with two time series prediction models respectively, Short term Memory Artificial Neural Network (LSTM) and Gated Loop Unit (GRU), and then the two combined models are used for two different network traffic data sets provided by an operator to predict the network equipment traffic. The experimental results show that the prediction models based on local context information enhanced attention mechanism have better prediction effect.