In order to improve the analysis effect of college students' sports and health data, this paper combines convolutional neural network algorithm to analyze college students' sports and health data, and explores the correlation between college students' sports and health. Moreover, this paper proposes a new dimension expansion method, which uses the least squares principle and the Kronecker product method to transform the algorithm into a data-driven way. In addition, this paper uses the fusion of time-frequency data and time[1]distance data as the input data of the network, and designs a CNN-LSTM network to automatically extract the movement behavior characteristics of college students contained in the spectral data.Finally, through data analysis, this paper verifies that the college student sports and health data analysis system based on the convolutional neural network algorithm has certain effects.