E-commerce recommendation algorithms are the hotspot of research currently. Relying on massive amounts of data, it is able to perform accurate modeling and prediction by using the deep learning algorithms, as well as enhance user activity and stickiness while bringing greater profits. In this paper, the author selects the Amazon data set as an analysis object, establishes a deep learning recommendation algorithm model based on DNN, and divides the data into training sets, verification sets, and test sets to predict product attributes. Experiments show that the deep learning recommendation algorithm based on DNN is superior to the traditional recommendation algorithms.