Research On Blog Post Denoising And Comment Recognition Method In Microblog Precision Marketing

With the rapid development of the Internet,social media has become an indispensable part of people's daily life.In these social media micro-blog with its convenient operation,information dissemination fast,real-time sharing and other features has gradually become the main battlefield of network marketing.At present,based on microblogging network marketing still stay in the traditional way of advertising and marketing,to promote business as the center,pay attention to the amount of advertising tasks completed,ignore the quality of advertising tasks,seriously affecting the marketing effectiveness of the enterprise.In order to solve the above problems,microblogging precision marketing aimed at the user as the core through the marketing tasks and microblogging users microblogging data analysis,select the appropriate microblogging users as a propagator to promote the task,and the use of reward incentive system feedback To help promote the user,in this way for the needs of users to provide accurate promotion services to improve the marketing effectiveness of enterprises.However,due to Sina microblogging information release gate low,platform management is loose,a large number of analysis tasks do not have the role of "invalid" data flooded them,making the microblogging data based on the accuracy of the analysis of the task.So how to remove these "invalid" data,leaving"valid" data for analysis is a key issue.According to the micro-blog data,In order to remove the advertising blog posts in the blog,the paper first constructs the text feature vector and the manually defined feature vector,and uses the stacked denoising autoencoders to process the two feature vector and obtain the processing.And then the two feature vectors are combined to obtain the third feature vector.The three kinds of feature vectors are used in the training of the maximum entropy classification model.According to the experimental results to find the best classification model,using the model of post processing to remove the advertising blog,the maximum entropy classification model experiments show that the P,R and F can reach 65.58%,87.9%,75.12%,Can effectively identify the vast majority of blog advertising.According to the review data,In order to solve the shortcomings of the previous review identification methods in the selection of the reference and individual differences,in this paper,we first construct different models for different bloggers,select the review to be replied by blogger as high quality reviews.A method based on the maximum entropy is proposed,first by the crawler and the word vector feature extraction,the feature selection method based on Wrapper is used to train the classification model according to the result of feature selection,and the validity of the proposed model is verified by the test data.The experimental results show that the proposed model is widely applicable to different bloggers,and the average accuracy,recall and F value of the classification can reach 66.64%,86.33%,and 75.2%.Finally,based on the above theory,the paper designs and tests the data preprocessing module of micro-blog precision marketing platform,namely,the blog denoising and comment recognition subsystem,to help the platform to make more accurate analysis results.