With the advent of the era of big data, the demand for mathematical modeling is increasing day by day, and traditional technologies face certain challenges when processing large-scale data. However, the large amount of data with low accuracy and high error of prediction is a challenging factor. So it is necessary to develop the respective advantages of big data technology in neural network algorithm combined with mathematical model. The tradition method linear regression is chosen to compared in this paper as a reference algorithm. At the same time, convolutional neural network (CNN) based on neural network algorithm is used as the experimental group algorithm. In the experiment, the steps of data preprocessing, grouping and randomization, and parameter setting were carried out, and the mean square error (MSE) was used as the evaluation index for comparison. Through experimental research, the mean square error of convolutional neural network technology is between 0.01 and 0.04.