Prediction of Mechanical Properties of Polymer Materials Using Extreme Gradient Boosting on High Molecular Weight Polymers

Abstract

The development of a new polymer materials depends on properties and it typically involves synthesizing a lot of compounds that finally becomes a new materials. This process is very expensive and take a long time. Therefore, the accurately prediction of the mechanical properties of polymer materials are required for understanding the rationale behind those predictions would be a valuable knowledge to the material studies. This work proposed the extreme gradient boosting to predict the mechanical properties of polymer materials. The model were developed, learn and the existing mechanical properties dataset. Then, the models were used to predict the mechanical properties for polymer materials. The predicted mechanical properties values were compared with the laboratory experimental results to validate the models. The experiment results have shown that the model could significantly enhance to predict the mechanical properties of polymer materials. In which way, this work could solve the problem of the polymer material and improve the efficiency of prediction obviously. This work evaluated the model with the logarithmic loss function, and the random forest classifier is used to test the model with all the features. Furthermore, this study enhance to reduce both time and cost of polymer material development. Even though, the accuracy of the predictions are not high enough, it has caused the insufficient number of sample in the dataset and the collection of data is widely scattered.