Spatial–temporal attention DBN and SBL for mobile network traffic prediction

Correct prediction of mobile network traffic is an important precondition for the rational utilization of network resources. This research develops a mobile network traffic prediction model based on the combination of Spatial–Temporal Attention, Deep Belief Network, and Sparse Bayesian Learning. The model first preprocesses mobile network traffic data, then extracts key features through a feature extraction method, and finally achieves accurate traffic prediction through a prediction model. According to the experiments, the average error of the model is just 1.99 GB/h after the iteration process, the root mean square error is just 2.51 GB/h, and the Coefficient of Determination is 0.924. As for practical usage, the error rates of the model at four different stages are 5.02%, 4.03%, 4.15%, and 3.86%, and the calculation time of the model is just 12.7 ms for a data size of 10,000, with a peak capturing ratio of 94.8%. Compared with baseline models, the model achieves significant improvement in both prediction accuracy and efficiency. The model provides an efficient and accurate solution for mobile network traffic prediction, and it supports network operators in resource scheduling, load balancing, and network planning, promoting more intelligent and efficient operation of mobile networks.