Sequential recommendation is a technique used to predict a user’s next purchased items by modeling their historical behaviors. However, traditional sequential recommender methods typically capture only sequential patterns and without taking other temporal information into account. As a result, their performance may be limited. In this paper, a time-aware sequential recommendation model based on attention mechanism (named TASR) is proposed. To enhance the modeling of user preferences and improve recommendation performance, our model captures temporal features of the sequence and combines them with item representations. Specifically, our approach utilizes an aspect-aware convolution to extract user and item representations from the embedding matrix. These representations are then merged with the temporal features during the temporal dynamic modeling process. Finally, multi-layer attention is used to derive users’ short-term preferences and the ratings of items are predicted using our TASR model. Extensive experiments on three datasets demonstrate the superiority of TASR against several state-of-art baselines.