Prediction Method of Telecommunication Traffic Data Based on Spatiotemporal Feature Fusion

With the rapid growth of the telecom industry, massive traffic data provides operators with an opportunity to gain insight into user behavior and needs. Effective traffic prediction helps carriers allocate network resources, including bandwidth, caching, and load balancing, improving network efficiency and performance and reducing resource waste. The purpose of this paper is to solve the problem of accuracy and real-time performance of telecom traffic data prediction. This telecom traffic data prediction method based on spatio-temporal feature fusion can improve the accuracy and real-time performance of prediction. Time domain feature extraction Bi-LSTM with negative feedback structure. In this way, more complex patterns and dependencies are introduced when processing sequence data, which further reduces network congestion and packet loss, and improves network service quality. Spatial domain correlation feature extraction a multi-scale and multi-head attention mechanism designed to better capture the dependencies of input sequences over different granularity or time spans. The experimental results show that the proposed framework has good capacity of traffic prediction