Research on the Fractal Characteristics of Online Public Opinion Risk Evolution in Mass Incidents

ABSTRACT Based on fractal theory and from the perspective of public opinion information, this study constructs a three‐dimensional identification framework encompassing ‘nonlinear characteristics, long‐term correlation, and scale‐free properties' to investigate the intrinsic dynamic mechanisms underlying the risk evolution of online public opinion in mass incidents. By collecting full lifecycle data from platforms such as Weibo for 19 typical incidents between 2018 and 2025, a time series of comprehensive risk measurement indices centred on ‘event energy value' was established. Nonlinear tests (Gaussian distribution fitting and Jarque‐Bera test), long‐term correlation tests (R/S analysis and Hurst exponent calculation), and scale‐free interval identification (random walk and DFA methods) were sequentially conducted. The results indicate that the risk evolution of online public opinion in mass incidents exhibits significant fractal characteristics, with the series showing a ‘sharp peak and heavy tail' distribution, persistent memory effects (H > 0.5), and statistical self‐similarity (presence of scale‐free intervals). The fractal characteristics vary across different types of incidents, with rights‐defence incidents displaying the most stable structure. The study verifies the emergence of macroscopic order from microscopic nonlinear interactions in the risk evolution of online public opinion in mass incidents, providing a theoretical basis and methodological support for predicting overall trends from local data and establishing dynamic early warning mechanisms.