Analysis And Early Warning Technology Research Based On Affective Computing In Online Community

The online community has the characteristics of anonymity,convenience and extremely strong timeliness,it is easy to produce public opinion.The spread of bad comments will disturb social order and threaten the public safety.Timely analysis and early warning of Internet public opinion can help relevant staff make timely and accurate judgments,prevent the spread of bad comments effectively,protect public safety and maintain social stability.Due to the text in online community has the characteristics of big data,spread fast,colloquial and strong emotional.The analysis and early warning of public opinion on the Internet not only needs to consider the emotional of the text but also needs to ensure near-real-time processing.However,the current research on public opinion early warning technology with Affective Computing is not perfect yet,and the real-time technology of public opinion analysis and early warning technology also needs further exploration.In view of the above problems,the public opinion analysis and early warning technology based on Affective Computing builds network public opinion knowledge.In order to ensure near-real-time processing efficiency,SparkStreaming which is a big data processing framework is used in the processing of text streams in online community.Combined with the emotional characteristics of the text,the natural language processing technology is used for text preprocessing.The emotional sentiment of the text is analyzed and the key words are obtained.The emotional calculation model is established by using the keyword as the core to calculate the text public opinion index,identify the text public opinion categories,and the text public opinion level.The experiment obtains the data of college forums and public forums to verify the effect from three aspects which are technical method,technical effect and processing speed.The experiment results show that the affective computing model improved the effect greatly.Compared with recent researches in related fields,it gain of 1.24%in F-measure over the"event-specific detection" proposed by F Laylavi et al.and outperforms emergency detect using contextual semantics by nearly 7.1%in F-measure.What's more,the Real-time early warning is implemented.The analysis and early warning technology research based on Affective Computing in online community has been applied to the network public opinion system and provided strong support for public opinion analysis and early warning of the online community.