The invention discloses an important event trend prediction method based on HC-TC-LDA. The important event trend prediction method includes the steps: firstly, adopting a web crawler to crawl specialnews data; then, extracting features through a topic model, and designing a topic model HC-TC-LDA fusing hierarchical clustering, word contribution degree and LDA for the problems that an LDA topic model has a large number of noise words and the number of topics needs to be manually set according to experience; and finally, constructing an important event trend prediction model by adopting a machine learning classification model. The prediction performance of the important event trend prediction method is superior to that of a traditional method which depends on expert knowledge to construct characteristic indexes. The important event trend prediction method can effectively realize high-precision prediction of the trend of the important event, has certain advantages in universality and timeliness, does not need more domain expert knowledge, provides certain auxiliary support for strategic decision making, and has a wide application prospect.