Research On Information Push Algorithm Based On Mixed Push Mode

With the emergence of the era of network fragmentation,information push has become a popular research in the field of computer.In order to solve the problems of sparse user interest matrix and cold start of new goods in traditional collaborative filtering algorithm,an information push algorithm based on hybrid push mode is proposed in this paper.The central idea of the algorithm is to use the BP neural network to integrate the two cooperative filtering push algorithms,so as to complement each other,and the push result is more accurate.And in the process of pushing,the ontology is introduced to correlate with user's interest words through the relationship between the concept hierarchy and the concepts,and then expand the meaning of user's interest words and make the recommendation more abundant.The hybrid push algorithm proposed in this paper is divided into three important parts:the first is the construction of user interest model.The user interest model is an accurate description of the user's hobby and an important link in the process of service push.Considering that the user's interest model is not immutable,the interest state parameters based on time are added to the vector space-based model representation to ensure that the constructed interest model accurately reflects the user's current interests.The second step is the combination of collaborative filtering push algorithm and ontology.In the traditional collaborative filtering push process,ontology is introduced to extend the meaning of user's interest words,which enriches the content of personalized push.Finally,the fusion of BP neural network is carried out.The results obtained by the two cooperative filtering push algorithms are taken as the input of BP neural network,and the final push results are obtained by neural network prediction.MAE and F1 are used as evaluation indexes in the experiment.The experimental results show that the hybrid push algorithm proposed in this paper effectively solves the problem of sparse user interest matrix in the user-based collaborative filtering push algorithm and the cold start problem of goods in the project-based collaborative filtering push algorithm.