As the Internet has grown more mature, various internet giants have also made huge profits through advertising. Image-based advertising is one of the most important forms of advertising. This study will use the CART decision tree as a research method, analyze and study the various features in the picture as variables, and then use RMSE, MAPE and MAD error indicators to examine the prediction ability of this model in this issue. The results show various variables. The degree of importance, this result allows the industry to follow in the production of pictures.