The random forest method in data analysis is used to study the optimization of impact properties for PP/ POE/

β

-Na/ATP blends. The model parameters are adjusted by learning and training the present 8 groups samples

and then the component particle size of these samples is lowered to refine the content of each component within the range of contents for these samples. 1750 samples are selected to verify the best two groups of predicted data. The experimental results are in good agreement with the predicted values. Thus

the better formula can be selected with less experimental times. The data flux of experimental verification is greatly expanded. The cost and the time of material formulation optimization and selection are effectively saved. Furthermore

the numerical analysis method is extended to analyze the orientation of PVDF/BaTiO

3

composites during the uniaxial tensile. The predicted results are basically consistent with the experimental results

which verifies the universality of Python numerical analysis method in the formula optimization of polymer blends.