Existing sequential recommendation models fail to consider both noise interference and the utilization rate of attribute information when integrating attribute information, which prevents them from effectively fusing attribute information and ultimately affects the performance of the recommendations. To address this problem, this paper proposed a sequential recommendation model based on multi-gated network fusion of attribute information. The model improves the utilization of attribute information by considering the value vectors of attribute information in the attention mechanism. Additionally, it introduces a gating mechanism into the feedforward neural network to reduce the impact of noise. Firstly, it generates item representations and their corresponding attribute information representations from the user's historical interaction sequence through an embedding layer. Then, it applies a decoupled representation fusion attention mechanism to generate a fused representation that incorporates attribute information from multiple representations. Next, it filters noise from the mixed representation using multiple gating linear structures in the multi-gated feedforward neural network to extract a more accurate mixed representation. Finally, it derives the user's final representation from the mixed representation, which is used for item prediction and attribute information prediction tasks to enhance the model's recommendation performance. The model is tested on three public datasets: Beauty, Sports, and Toys. Compared to well-performing baseline models, it achieves improvements of 1.40%, 1.76%, and 1.24% on the Recall@20 metric, and 3.50%, 3.48%, and 2.01% on the NDCG@20 metric, respectively. Experimental results indicate that the proposed model effectively reduces noise interference and enhances the utilization of attribute information, thereby significantly improving recommendation performance.