Research And Application Of Personalized Recommendation Algorithm Based On Probability Matrix Factorization Model

With the rapid development of the Internet,a large number of social networks have appeared,and the data of users and goods on the network has increased dramatically.At the same time,people are more inclined to recommend relevant information based on personal interests.However,in the current research,model adaptability and coping with data sparse scenarios are still relatively difficult problems to solve.This article mainly researches personalized recommendation.By constructing a social network personalized recommendation model,combining feature extraction and big data mining methods,it can realize the information mining in the social network recommendation process and improve the system's recommendation performance.The main research work of this paper is as follows:(1)Aiming at the factors that the traditional probability matrix decomposition model is not adaptive enough,a personalized recommendation algorithm is proposed based on the observed user's action characteristics.Create an Internet social user behavior information feature mining model,combine fuzzy information perception methods for information fusion processing in the personal recommendation process of social networks,and construct a hybrid recommendation model of social network user preference features under the control of association rules to achieve user preferences Feature mining,personalized information recommendation of social networks based on user behavior preferences(2)In related popular applications,deep learning algorithms play an important role.In industry,deep learning can improve the efficiency of the original algorithm.The steps of constructing features in the recommendation process,generating recommendation candidate sets,and predicting recommendation scores have good application effects.Combining recommendation algorithms with deep learning algorithms has become a research hotspot in recent years.The recommendation model for the probabilistic matrix decomposition model has data sparseness defects.Design a neural network algorithm combined with deep learning and a probability matrix decomposition model to make full use of features in high-dimensional sparse scenes to make accurate recommendations for users.(3)In reality,a single user does not have many records in the social network,that is,the data is sparse,which will lead to a lower accuracy of the user weight value.To solve this problem,a probability matrix decomposition model based on multiple indicators is used.Multi-index scoring judges the user's personal interests and hobbies based on the review context information,and recommends more accurate information to the user based on this.In the multi-index scoring system,the overall user's score is used to influence the weight.Multi-index scoring has a greater impact on matrix decomposition.When constructing the matrix decomposition model,we must fully consider the relationship between each index.Users,projects and other related information are integrated into the user-project-multi-index relationship model to improve the recommendation effect.