Research On The Evolution Analysis Of Online Social Network Sentiment Based On Big Data

With the rapid development of the Internet and the rise of the mobile Internet, online social network has penetrated into several fields of people’s life. Extracting knowledge from the massive Internet data and making this data a better application is one of the popular research subjects. The network information dissemination shows several characters as a wide range, fast speed, and regularly, particularly in a social network. Public opinion analysis based on the prior social network is one of the hot spots in the current research. We conducted the experimental analysis with actual data, and simulation of evolution and construction of emotional social network propagation model Based on the research of the social network structure. Our study makes the following contributions:1) Emotional analysis of a large number of number data extracted for SNS platform. Firstly, we extract a lot of semistructured data From current typical social network platform such as Baidu Post Bar, the micro-blog, and so on. Secondly, we classify the sentences to the sentiment polarity using the feature subset of the feature words extracted from the pre-processed data. Finally, we compare the results of the experiment.2) This study we focuses on the applications of network structure and transmission of information in the network. Firstly, we study the algorithm of construction and growth of network structure based on the social network topology structure. Secondly, We propose three dissemination strategies for the Mood spread of network algorithm. The model created considers the influences of the network structure, information sources, the relationship, and personal character. We also conduct experiments by adjusting the parameter.3) Based on the analysis on trial data. There are 5 stages in the spread of information in the network. We can achieve the purpos of controlling the spread of information by adjusting the characteristics of network because of the difference trend by adjusting the parameter in each stages.4) Finally, we develop an interactive platform that allows users to directly observe the clustering results.