Online News Credibility Evaluation with Bayes Model

Abstract

In today's Internet information explosion, all kinds of information floods in people's daily life. Meanwhile, the credibility of news has become one of the key issues that people pay attention to. In order to realize the evaluation of news credibility, this paper mainly uses k-means clustering method to carry out quantitative analysis on the training set data. After defining and judging indicators of different attributes, it carries out binomial distribution mapping of quantized attribute indicators, taking 0 or 1 as the parameters of direct evaluation indicators. The optimized Bayes formula is used to calculate the probability of news credibility corresponding to different indicators, and the prior probability and posterior probability of news credibility of each attribute are obtained. After the evaluation, the appropriate probability is selected to represent. Finally, the sum of the confidence probabilities corresponding to the binomial distribution mapping of each index in each article is calculated and multiplied with the correction parameters to obtain the final quantitative evaluation result of network news credibility. In this paper, a mathematical model is established based on the given data set. The model score is compared with the expert score, and the accuracy of the evaluation of the authenticity of each piece of news is evaluated by the residual squared and response model. Based on the established indicator framework of online news credibility evaluation, some Suggestions and Suggestions on how to improve news credibility are proposed to news publishers or social media.