In the face of the problem of low accuracy of university sports load data prediction method, a deep learning university sports load data prediction method is designed. Identify the style and rules of human movement, extract the characteristics of time domain to calculate in frequency domain, construct the target tracking model by deep learning, calculate the error of the output layer, extract the characteristics of college sports load, judge the rationality of the movement contact configuration between bones, and design the data prediction method. Experimental results: The average prediction accuracy of the college sports load data prediction method in this paper and the other two methods are 0.417, 0.342 and 0.333 respectively, indicating that the precision of the college sports load data prediction method designed after the full integration of deep learning technology has been improved.