With the deepening of the new curriculum reform, physical education is receiving more and more attention from the state and schools. Aiming at the learning data of physical education students, a physical education learning behavior recognition model based on Adaboost-BP neural network is proposed. The BP neural network is used as a weak predictor in MapReduce environment, and a strong predictor is constructed by the Adaboost algorithm combining the results of the weak predictor to recognize the physical education learning behavior. On this basis, the physical education teaching strategy is adjusted, and physical education stratified teaching based on the Adaboost-BP model is proposed, and the teaching experiment is implemented to evaluate its application effect.The Adaboost-BP model has a good effect of recognizing physical education learning behaviors, and the average error of the recognition error has an absolute value of 0.029 and a relative value of 3.79%, which is smaller than the comparative method.The model will The model identifies the physical education learning behaviors of 100 students into three categories: “excellent”, “moderate” and “poor”. After the adjustment of teaching strategies, the three categories of students were improved in all indicators of physical fitness, and in terms of course enjoyment, motivation, skill mastery, course content rationality, learning experience, etc., the Adaboost-BP model can realize the stratification of students’ behaviors, and then guide the adjustment of learning behaviors and physical education courses, and promote the improvement of physical education teaching quality.
