Abstract: 【Purpose/significance】Conduct research on multi-stage public opinion that has sustained effects or even serious conse?
quences, try to establish a fermentation early warning model based on probabilistic analysis methods to accurately diagnose the cause
of fermentation, and hope to provide decision-making basis for network public opinion governance and control.【Method/process】Ab?
sorb the experience and lessons from traditional models, reduce subjective evaluation indicators, increase the degree of refinement of
data layer indicators, and use Bayesian probability to construct a fermentation prediction model. At the same time, accurate diagnosis
of the cause of fermentation is carried out through the principle of the greatest possible explanation.【Result/conclusion】Through us?
ing 55 of the 60 multi-stage cases and 27 of the 30 single-stage cases as training data to construct a multi-stage early warning model
and using the remaining 5 multi-stage and 3 single-stage cases as a test group, the prediction result of the fermentation trend obtained
by the test is consistent with the facts.【Innovation/limitation】The internal causes of multi-stage fermentation and the multi-stage di?
agnosis of the causes has been explored, the accurate grasp and scientific quantification of predictive indicators has been achieved,
and the useful theoretical support for early warning of online public opinion and the formulation of intervention measures has been pro?
vided.