Comparison of Machine Learning Approaches for Prediction of the GlassTransition Temperature of Polymers based on a Chemical model

Glass transition is a very important phenomena in polymer science which helps to understandpolymers behavior at different temperatures. Modeling and prediction of the temperature that thisphenomena happens, also known as glass transition temperature, has always been and important partof physical chemistry of polymers. Two different machine learning algorithms, linear regression andrandom forest have been implemented in this study in order to predict the glass transition of polymersbased on the important structures and groups on their repeating unit. Both models showed promisingresults with coefficient of determination of around ۰.۹, this was while the random forest modelshowed great results for some structures and very bad results for others while linear regressionshowed average to good results for most predictions and the average error for both models wasapproximately ۱۰%.

کلیدواژه ها:

Polymer Characterization ، Property Prediction ، Machine learning و linear regression ، glass transition