Food safety problem has been a hot topic of national concern, which related to many areas of society. In order to know hot issues that relate to food safety in timely, food safety hot topics and other hot topics of the similarities and differences in detection methods were compared. The food safety surveillance topic detection model was constructed and the clustering algorithm was used for text mining for food safety data to achieve the topic detection. Through the experimental results, the evaluation of the Single-Pass algorithm was better than the K-Means algorithm, which could effectively detect food safety topics.