The sequential recommendation predicts the items that a user may interact in the near future via modeling the user behavior sequences. However, the existing methods adopt shallow networks which cannot extract deep features, and can't fully integrate the user general preferences. We proposed a model named a sequential recommendation model with convolutional neural network and multiple features (SRM), which uses multi-layer convolutional neural network to capture both the point-to-point and union-type sequential patterns. In addition, the model adds the user general preferences and item attributes to improve feature diversity. The experiments on public dataset proved that the SRM model can effectively predict the user's future actions, and the performance of the model is better than the existing sequential recommendation models.