Quantifying The Evolutionary Trends Of Online Public Opinion Via Incorporating Machine Learning Strategies

The rapid development of big data,artificial intelligence,5G and other emerging information technologies has made social networks an increasingly popular medium for instant access,rapid dissemination and simultaneous sharing of information.However,due to the high degree of uncertainty,proliferation and instability of online opinion generated by the influx of internet users into social networks,it has a strong tendency to lead and influence the trend of public opinion.Therefore,research related to the issue of online public opinion has become an important focus of scholars at home and abroad at present.At the present stage,the existing research on online public opinion is still characterized by a predominantly qualitative nature,a lack of quantification and an imperfect methodological mechanism,which is actually difficult to meet the realistic needs of relevant government departments in the information age for analysis and supervision of online public opinion.How to grasp and predict the overall trend of online public opinion,how to open up new research perspectives to interpret the trend trajectory of online public opinion,and how to construct effective models for the quantitative presentation of online public opinion,so as to effectively improve the practical effects of controlling,regulating and guiding online public opinion,are urgent needs that need to be addressed in the current context.Based on the perspective of information behavior,this paper uses literature analysis,empirical research and machine learning methods,and combines information dissemination theory and information demand theory to construct a secondary topic identification model of online public opinion and a diffusion degree model of public opinion.Chapters 3 and 4 of this study are the core theoretical chapters and Chapter 5 is the empirical analysis chapter.Chapter 3 deals with the secondary identification of online opinion themes.Firstly,the improved LDA model is used to carry out the primary identification process of online opinion hot topics,and secondly,in order to further improve the identification accuracy,the secondary identification process of topics is carried out by the third-order nearest neighbour propagation clustering algorithm.This chapter provides the basis for the construction of the subsequent online opinion diffusion degree model.Chapter 4 provides the construction of the online opinion diffusion degree model.The CNN-BiLSTM neural network was first introduced to construct a sentiment element model to obtain a classification of users’ sentiment tendency and a reasonable sentiment polarity score,and on this basis,the coefficient of variation theory and information entropy theory were combined to construct an opinion diffusion degree model to measure the degree of possible diffusion during the evolution of public opinion.Chapter 5 is an empirical study.Based on the life-cycle theory,this paper studies the hot topics of online public opinion at different stages of the life-cycle and the extent of public opinion spreading in the process of evolution,and compares the trend of the Baidu search index of the topic to prove the validity of the proposed model.At the theoretical level,this paper constructs an opinion diffusion degree model,which provides certain theoretical support for public opinion management;at the practical level,the public opinion supervision department is able to carry out practical control in a quantitative manner through specific numerical values,which in turn can effectively guarantee the healthy and sustainable development of online public opinion.