In order to improve the ability of analyzing and evaluating the behavior of phased independent physical exercise, this paper puts forward a prediction method based on Bayesian analysis. It constructs the statistical time sequence analysis model of phased independent physical exercise behavior prediction with the big data feature detection method to explore and extract the big data of physical exercise behavior. Based on the Bayesian analysis and prediction idea, orderly clustering of the behavior statistical feature sequence is carried out, and the information clustering and categorizing in the process of physical exercise behavior prediction are combined with fuzzy C-means clustering analysis method in this paper.Characteristic quantity of associated rule for the statistical time sequence is extracted, and the phased independent physical exercise behavior is accurately predicted in the weighted Markov chain. The simulation results show that the accuracy of this method is high, and the ability of quantitative analysis of autonomous physical exercisehas been improved.