To address the challenges of urban traffic congestion management, this study proposes a traffic state identification and prediction method. First, a traffic state identification model was constructed using the Fuzzy C-Means (FCM) clustering algorithm based on congestion identification factors such as traffic flow, average speed, and average headway. Second, a time-series prediction model for traffic flow parameters was established, enabling traffic state prediction by inputting the predicted parameters into the FCM. The prediction model, based on Gated Recurrent Unit (GRU) incorporates a bidirectional GRU to fully exploit data features. To further capture the influence of different time steps, the model introduces a temporal attention mechanism to focus on critical time steps and is optimized using a genetic algorithm. Finally, vehicle operation parameters were extracted to construct training and testing datasets, validating the model’s traffic state prediction accuracy using video data from the intersection of Yingbin Avenue and Hongye South Road in Dongguan City. The results demonstrate that compared to the pre-optimized GRU, the proposed model improves traffic state prediction accuracy by 4.9%, achieves the accuracy of traffic state prediction of 82.6%, and effectively distinguishes traffic flow state levels, enabling short-term traffic state prediction.