Analysis Of Online News Popularity Based On Ensemble Learning

Nowadays,the Internet is the same as a spider web,connecting all the world.All kinds of information in the world can be transmitted and feedback rapidly by the online.The validity of online news is the number of times it is viewed.In order to promote their news on the Internet,many editors often measure their news content,form,release time,links,and so on repeatedly.The most concerned problem for editors is how to maximally promote online news.This article explores the main factors that affect the popularity of online news,based on the number of news promoted on Mashable which is a Internet blog.Firstly,the possible impact factors were selected by principal component analysis to find the main factors of news popularity.Keep these factors as affect factors about the research questions of news popularity.Secondly,the number of popularity of news is viewed as the target variable and influence factors as independent variables.Variables were analyzed by multiple linear regression analysis,logistic regression analysis,K-nearest neighbor and support vector machine analysis,and the corresponding fitted model values and scores were obtained.But these scores are less than 1,it can be seen that the result of using one model alone is not satisfactory.Therefore,ensemble learning method is chosen in this article that use every model as a learner and use the proportion of score of every model to the total scores of models as a weight.By testing,we can see that the result of ensemble learning model is good,and the number of popularity of online news is predicted precisely.The model of ensemble learning improves the generalization ability.Lastly,according to the result of the ensemble learning,effective measures are proposed for news editors to achieve the purpose of improving efficiency and increasing the popularity of online news.