A large number of users are getting news through online news media websites,and in the face of the huge amount of news data,users are often caught in a selection dilemma.At this time,media need to consider how to provide users with personalized recommendations accurately and efficiently.The main task of this study is to use the characteristics of news and users' history to provide users with accurate news recommendations in order to improve user satisfaction and system performance.(1)In order to solve the cold start problem of users,for new users to make recommendations,and for news front page headline news setting.In this paper,we propose a recommendation model with the highest future impact based on news features,which uses the ordinary least squares model to predict the future impact of news using the past impact,news categories and other features,and then generates a list of user recommendations after ranking.(2)In order to calculate users' preferences and personalize recommendations to them according to their browsing history.In this paper,a user preference-based deep knowledge-aware network model(UP-DKN)is proposed for click-through rate prediction.Different weights are assigned to different news items in the user's reading history.UP-DKN mainly utilizes a knowledge-aware convolutional neural network(KCNN),which implements a network with word vector entity vectors as well as weight vector alignment.(3)In order to solve the user preference shifting problem,in this paper,we propose a sequential hierarchical attention network recommendation model(IFSHAN)based on item features.Specifically,the first attention layer learns users' long-term preferences based on historical browsing records,and the second layer outputs final user representations by coupling users' long-term and short-term preferences.(4)Based on the above three recommendation models,this paper designs and implements a prototype news recommender system,which can effectively solve the cold start problem of users and provide users with personalized recommendations,improve the accuracy of recommendations,and thus obtain a better user experience. |