Short Text Classification And Personalized Recommendation Based On Sina Microblog

With the rapid development and popularization of social networks,there are more and more registered users of Microblog.Microblog has become one of the most popular social media in China today.Therefore,the amount of information carried by Microblog is also growing.How to get useful information in huge amount of information and how to make accurate recommendations to users have become the focus of e-commerce.This paper does two aspects based on Sina Microblog.One is the short text classification based on Sina Microblog and the other aspect is the personalized recommendation based on the user interests model.In the short text classification model,our contributions mainly lie in the feature selection and parameter optimization.For feature selection,first,we filter the lexical terms that are related to the category labels to get the primary features set.Then,the primary features set is imported into the LDA topic model to further select features.For parameters of LDA topic model and kernel-Support Vector Machine(SVM),we choose the optimal parameters by K-fold cross-validation.Experiments show that our classification model has significant classification effect.In the personalized recommendation model,our contributions are reflected in the construction of the user interests model and reducing the sparsity of rating matrix.When building a user interests model,we use the quadratic mining method to obtain the followings' personal labels.In addition,we weight user interests using a weighted approach.Because the sparsity of the rating matrix used in the recommendation system directly affects the recommendation performance.So,this paper combines user-based and item-based collaborative filtering to fill sparse rating matrix.Experimental results prove that the improvement points of our personalized recommendation model all help to improve the recommendation performance.