A Hybrid User Behavior Prediction Model for Wireless Networks Based on Collaborative Filtering and GRU Networks

With the rapid growth in wireless network user scale and data complexity, efficiently predicting user behavior to optimize network resource allocation has become a critical challenge in the fields of communication and intelligent recommendation. This paper proposes a prediction model that integrates collaborative filtering with gated recurrent unit (GRU) networks, aiming to capture both behavioral similarity among users and temporal dependencies within user behavior sequences. An end-to-end joint learning framework is constructed, where an embedding-level fusion strategy is employed to effectively integrate collaborative and sequential features. Experiments are conducted on a dataset structured in alignment with typical wireless access logs from telecom operators. Results demonstrate that the proposed model achieves superior performance in both behavior classification and traffic estimation tasks. Specifically, the model attains a Precision@5 of 0.742 for behavior prediction and reduces RMSE for traffic estimation to 61.8 KB, outperforming both traditional approaches and existing deep models. Further validation through comparative and ablation experiments confirms the effectiveness of the key modules. This work presents an effective solution that integrates collaborative behavior modeling with sequence learning, offering theoretical support and practical insights for behavior-aware and service-optimized intelligent communication systems.