An Event-Driven Public Opinion Risk Early Warning Method for Autonomous Robotaxi Services
Published:
The present invention discloses an event-driven public opinion risk early warning method for autonomous robotaxi services, belonging to the technical field of data processing. The method comprises the following steps: constructing an event-stage variable, a relative event-time variable, and their interaction terms using the occurrence date of a target event as the temporal anchor, and incorporating them as document-level covariates to estimate topic proportions, sentiment discourse expressions, and covariate regression coefficients; based thereon, identifying risk-related topics, calculating the variation intensity of each risk topic before and after the event, the degree of sentiment shift, and the post-event trend variation indicator, and combining these with interaction popularity to generate a topic risk index; subsequently determining topic weights according to the proportion of each risk topic and the intensity of negative sentiment, and generating an overall public opinion risk index through weighted aggregation; and finally outputting a risk level, key risk topics, and an early warning result according to predefined thresholds. By identifying structural shifts in topics and sentiments following an event and transforming them into quantitative risk indices and graded warning outputs, the present invention achieves the conversion from dynamic public opinion analysis to risk early warning.