Research And Application Of News Influence Model Based On User Behavior Analysis

With the rapid development of Internet technology,the focus of the news media industry has shifted from traditional paper-based media to online media.Network news is the main source of the formation of network public opinion and social public opinion,it is particularly important to accurately calculate its influence.The existing models of news influence have some problems,such as incomplete indexes,artificial selection of parameters,and insufficient flexibility of models.In response to the above problems,this paper explores two perspectives to improve the accuracy of the model and consider the comprehensiveness of the perspective based on user behaviour analysis technology,and finally applies the model to the field of online hot news monitoring to provide useful information.The main work of this paper is as follows.1.Construct a model of news influence based on neural network.To address the problem that the current news influence model research is not accurate enough when quantifying news influence.Combined with user behaviour analysis,a model that uses a non-linear network structure instead of a linear weighted model to measure influence is proposed,and optimisation is made in terms of indicator processing,weight selection and elimination of the influence of the time factor.Using long-short term memory neural networks,improvements are made in data processing to avoid wastage due to missing values;an indicator processing algorithm is proposed to eliminate the interference caused by the time factor;a news data event segmentation method is designed to calculate influence in terms of events.A back-propagation neural network,which is good at solving non-linear computational problems,is used to construct a quantitative model of news influence,and the weights of each indicator are automatically selected based on historical data.2.Construct a news influence model based on improved PageRank.The following problems of the current news influence quantification model are addressed: the perspective considered in the quantification of influence is not comprehensive enough,and the influence calculation model does not conform to the actual news influence development law.Combining linear weighting models,neural network models and propagation relationship models,I propose a comprehensive approach to calculate news influence that involves information retrieval,user behaviour and news interactions.The goal is to achieve a more comprehensive and objective calculation of news influence.Using directed graph knowledge and improved PageRank algorithm formula,I propose a quantitative formula to calculate the increase and decrease value of news spread interaction.Combining a linear weighting model,a neural network model and a propagation relationship model,a quantitative model of news influence is proposed for comprehensive consideration.3.A hot news monitoring application based on the news influence model.To address the problem of poor quality and lack of effective display in the field of online hot news monitoring,combined with the data mining-based news influence model,to provide an application solution to extract high influence news from online news and recommend it to government departments.This programme can effectively increase the efficiency and quality of opinion monitoring and provide a perspective for decision makers to consider.