Abstract: 【Purpose/significance】In view of the existing topic evolution method is difficult to meet the demand of prediction purpose. This article to build knowledge topic evolution prediction model from the perspective of knowledge dynamic development，which pro⁃ vide a new way to explore the development and trends of science filed.【Method/process】We apply the LDA model to extract knowl⁃ edge topics, and build topic steady-state and topic heat evolution prediction model by using Markov model and hidden Markov model. 【Result/conclusion】With scientific literature in the field of cloud computing as the object of empirical analysis, the results show that our model can forecast the knowledge topic steady-state distribution and heat trend based on historical data, and the heat prediction accuracy is higher than that of grey model.【Innovation/limitation】This paper only considers the heat change within the topic in the hor⁃ izontal dimension, and does not compare the knowledge topics in the vertical dimension.