CMMC-Model

This is the code for 'Using Active Learning for the Computational Design of Polymer Molecular Weight Distributions'

package

• python==3.7.11
• numpy==1.21.6
• pandas==1.3.4
• scikit-learn==1.0.2
• modal==0.4.1
• We suggest that you can install these packages by pip.

CMMC Model

We provide two different versions CMMC model. CMMC only used Mn and PDI as the output, and the training data are mode0_data.csv and mode1_data.csv. In order to describe MWD more accurately, we added Skewness and Kurtosis to describe MWD in CMMC_v1, and the traning data for CMMC_V1 is data_final.xlsx.

SHAP

We also provide the code for SHAP Analysis, you can find it in SHAP_xgb.ipynb.

Note

1. Xgboost model performs better than random forest model when the Skewness and Kurtosis are added to describe the MWD. So we choose Xgboost model in CMMC_v1 and random forest model in CMMC.