Prediction of Public Opinion Event Types Combining Retrieval-Augmented Generation and Large Language Models

The prediction of public opinion events plays a crucial role in social governance, public safety, and enterprise management. Accurate prediction of event types helps identify potential crises, effectively guide public opinion, and provide scientific support for decision-makers. This study leverages large language models and introduces a Retrieval-Augmented Generation (RAG) system to propose a novel method for improving the accuracy of public opinion event prediction. Using a small amount of labeled data, the method retrieves relevant data slices to dynamically generate high-quality few-shot prompts, enabling the language model to classify events more accurately and enhance its predictive capabilities. The findings demonstrate that effective few-shot prompt generation significantly optimizes the application of RAG systems in public opinion analysis, offering a new perspective for event prediction and showcasing its potential for real-world applications.