With the rapid development of the Internet and information flow,the Internet has increasingly become the main information source for people.C Improving the efficiency of information browsing,accurately pushing personalized content to users who are interested in have become a hot demand at present.First,Python is used to obtain the one-week exposure history of users on a platform in information flow products,and the data is then processed and analyzed.The Transformer deep neural network model and the most similar user estimation model are introduced and combined together to predict the click rate and duration of users browsing each content.Finally,the model's explanatory nature is enhanced,and it is proved to be more sensitive to the preference of recommendation sequences in different orders.
recommendation algorithmTransformerneural networkthe most similar userssequence evaluation
NETL
NSTL
维普
万方数据
