Resume  

​

 Recently, accurate predictions using machine learning (ML) algorithms have been mentioned in several fields. In this study, we constructed two quantitative structure-property relationships (QSPR) models by a dataset of organic compounds (small molecules) having available experimental heat of vaporization ) data, to predict the heat of vaporization of polymer repeat units. For that, Multiple Linear Regression (MLR) and Kernel Ridge Regression (KRR) models were created. Thirteen workable parameters (descriptors) were involved in both ML models, including count descriptors and quantum chemical descriptors, and calculated by different approaches from both small molecules and repeat unit structure. The dataset of small molecules was randomly divided into a training sample (70%) and a testing sample (30%). MLR demonstrated the best performance for  outcome prediction for polymer repeat units with a correlation coefficient (R2 ) of 0.835, a MAE of 3.769 (kJ/mol), and a RMSE of 4.531 (kJ/mol). Our results indicated that most of the predicted  values are in good agreement with the experimental data.

​

Publications