The work aims to explore the machine learning-based effect evaluation model of print public service advertisement based on the basic theory of data science, and realize the design effect evaluation of print public service advertisement. Firstly, the image data set of print public service advertisement was segmented according to the ratio of 7∶3, and the characteristics of advertisement images were extracted from multiple dimensions. Then, the classification model was constructed by integrating random forest with Multi-layer Perceptron Model and XGBoost algorithm, and the classification evaluation model of the strong and week advertising effect was obtained through training. The test results showed that the F1 value in the training set was 1.0, and the F1 value in the test set was about 0.897 6, which could correctly evaluate the advertising effect. The multi-feature fusion and multi-model integration model based on machine learning can realize the accurate evaluation of print public service advertisements, and provide a new method for the evaluation of print public service advertisements effect.