Research And Implementation Of Personalized News Recommendation Algorithm Based On Improved LDA Topic Model

The mining of news text information has always been a research hotspot in recent years,and topic model analysis is one of the mainstream methods of news text information mining.Traditional topic models,such as LSA(Latent Semantic Analysis),PLSA(Probabilistic Latent Semantic Analysis)and LDA(Latent Dirichlet Allocation),mainly build user interest models by mining hidden topics in text information,but with text information The advantages of traditional models for mining topic information are gradually declining,which brings certain challenges to text mining;sentiment analysis uses subjective information such as user evaluations of text information to mine and analyze users' preferences for a certain news category,and sentiment analysis Mainly based on sentiment classification,the sentiment information of user comment text is divided and classified,and the classified data is used to analyze and predict the user's interest preferences.Based on the existing research,the dissertation proposes a recommendation algorithm that combines sentiment analysis,time factor and user trust with LDA topic model(hereinafter referred to as NR?ILDA).The main research work is as follows:(1)LDA topic modeling,through Gibbs Sampling(Gibbs sampling)obtain the posterior distribution of the hyperparameters,obtain the LDA model,and obtain the user-topic probability matrix U1 through multiple iterative experiments;(2)perform sentiment analysis on the comment information in the dataset and divide the text The part of speech of the words in the information,the time of user comments,and the trust factor between the user and the user are considered,and the user weight matrix U2 is calculated through the similarity calculation;(3)The matrix U1 and the matrix U2 are weighted to obtain the user preference matrix U,finally get the Top-N list,recommend it to the user,and then add the news information that the user has browsed to the original data set for iterative training and recommendation;(4)Perform experimental comparison and analysis of the NR?ILDA algorithm with other algorithms,and The algorithm is applied to the personalized news recommendation system.The experimental results show that the NR?ILDA algorithm is superior to other algorithms in terms of accuracy,recall and F1 value,and the recommendation quality is higher,which is feasible in the application of personalized news recommendation systems.