Public health emergencies are characterized by fast spreading speed, wide range of influence, and great social harm, and the traditional response method has the problems of monitoring blind area and response lag. With the development of big data and artificial intelligence technology, the intelligent detection and response system based on pattern recognition can realize the integrated management of the whole process, improve the early identification and efficient disposal of community public health events, and provide information support for precise prevention and control. In this study, the BiLSTM+CNN hybrid neural network model was used to extract the deep semantic features of the text, and fused the multidimensional information such as basic user attributes, behavioral features, text features and communication features to realize the intelligent recognition and processing of the information on public health emergencies. The experiments use CHECKED extended dataset for model validation, which contains 813 rumor texts and 1894 non-rumor texts. The results show that the proposed BiLSTM+CNN multifeature fusion model performs well in the rumor recognition task, with an F1 value of 0.985, and accuracy and precision of 0.978 and 0.974, respectively, which are better than the existing mainstream models. Further analysis of the response effect of community residents in Wuhan shows that the global Moran’s I index of sentiment index and case index is -0.115, showing a significant negative correlation in the period of strict prevention and control. The results of the study proved that the pattern recognition method based on multi-feature fusion can effectively improve the intelligent detection and response ability of public health emergencies, and provide new ideas for community public health emergency management.
