Research And Implement Of Text Mining And Intelligent Recommendation Based On Browsing History

The rapid development of the Internet has produced Web text with a large amount,complex structure and highly real-time processing requirements,the general text recommendation method depends on the attribute calculation or keyword matching,resulting in the recommendation is not ideal.In order to solve this problem,this paper extracts the textual information of the text in the user's browsing record,performs the feature representation and the semantic analysis of the text,and selects the web page text with high similarity to the user's browsing in the same category,combined with the user's interest to the text of the page,the sort results will be recommended to the user.Completed the research work in intelligent recommendation.The main work includes the following three aspects:(1)This paper presents a Convolution Neural Network(CNN)text model using Semantic Aware Pooling(SAP).Among them,semantic aware uses semantic perception processing module in the Hierarchical Network of Concepts(HNC)theory of sentence pattern,combined with Probability Context-Free Grammar(PCFG)method for sentence analysis,in order to improve the CNN pooling layer to deal with the convolution text feature.Using the improved model to simulate the text similarity and text semantic analysis,the feasibility and superiority are verified.(2)On the basis of text feature representation and calculation,this paper uses the long text similarity calculation method constructed in the related research of traditional Vector Space Model.To calculate the similarity between the user's browsing text and the candidate set text.Where the candidate text set is obtained by the user browsing the record and the recommended text clustering.And then collect the user page operation behavior and calculate the user's interest in the text of the page.This paper uses a method to combine the text similarity model with the user interest model to calculate the candidate text with the higher score of the similarity and interest value as the final Web text recommendation list.In this paper,the proposed Web text recommendation model is experimentally analyzed.Experiments show that this method can speed up the convergence rate of the algorithm and reduce the Mean Absolute Error.(3)The methods involved in this paper are integrated and designed.The intelligent recommendation tool is implemented on the BS architecture,and completed the tool's deployment and testing.All the functional testing and performance testing has met the basic requirements.