Sequential recommendation has gained significant attention in the realm of recommender systems, especially with the recent advancements in artificial intelligence. Many researchers have been integrating deep learning techniques to enhance the effectiveness of sequential recommendation. However, many existing methods rely on modeling with specific Item ID, which inevitably overlooks the inherent attribute features of the items and resulting in challenges such as the long-tail effect and cold start problems. In this paper, we propose a novel sequential recommendation model named MASRec, which integrates multiple attribute information from items. Our approach divides an interaction sequence into multiple attribute sequences, enabling independent learning of user interests on each attribute. Furthermore, we introduce a weight parameter for each attribute to capture user preferences across different attributes. The prediction for the next item is derived by aggregating and weighting the prediction results from each attribute. Finally, we have conducted experiments on four real-world datasets with MASRec, and the results demonstrate its ability to significantly enhance recommendation effectiveness, better than mainstream baseline models currently available.