Quantitative structure-property relationship (QSPR) study to predict retention time of polycyclic aromatic hydrocarbons using the random forest and artificial neural network methods

Abstract

In this study, a quantitative structure-property relationship (QSPR) was used based on powerful methods, including random forest (RF) and feed-forward back-propagation neural network (FFBP-NN). These methods were carried out for modeling and prediction of retention time (RT) values of polycyclic aromatic hydrocarbons (PAHs) compounds. Also, the number of trees (nt) and the number of randomly selected variables to split each node (m) were studied as effective factors in RF method. The best results were obtained at nt = 300 and m = 7. In addition, statistical parameters were investigated, and the correlation coefficients (R2) of the RF (m = 7), stepwise regression artificial neural network (SR-ANN), and RF-ANN models were 0.9954, 0.9920, and 0.9831, respectively. The obtained results show that the RF model has better performance compared to the other models. So, it can be used as a powerful tool for QSPR studies.