Research on Early Warning System of New Media Events Based on Model Segmentation and Feature Integration

In recent years, with the increasing number of Internet users and mobile phone users in China, various social contradictions have been discussed locally from the real world to the global discussion of the virtual world. In order to distinguish new media events in time and accurately, based on the analysis of common clustering algorithms, a hybrid K-means genetic algorithm suitable for clustering new media events is proposed by combining genetic algorithm with K-means algorithm. The algorithm uses the optimal preservation strategy, single-point crossover and single-point mutation to ensure the convergence of the hybrid K-means genetic algorithm to a greater extent. Multi-round merging based on dynamic weights is adopted to make segmentation results suitable for retrieval requirements, and the retrieval method of feature integration is improved. On the basis of the initial weights, the weight knowledge base can be stabilized through a certain number of user feedback training processes. Finally, according to the weights in the knowledge base, different features are integrated for retrieval. Experimental results show that the algorithm is effective.