Hybrid data-intelligence algorithms for the simulation of thymoquinone in HPLC method development

In this study, two different single non-linear [feedforward neural network (FFNN) and support vector machine (SVM)] models with a traditional linear regression model [multi-linear regression (MLR)] were employed to predict the qualitative behaviour of thymoquinone (TQ) in HPLC method development interms of retention properties. The simulation involves the use of the concentration of the standard, the mobile phase, pH and flow rate as the corresponding input variables, while the retention time (tR) of TQ is considered as the dependent variable. Four performance indices were employed to determine the accuracy of the models, namely correlation coefficient (R), root mean square error (RMSE), mean square error (MSE) and determination coefficient (R²). Subsequently, hybrid models were proposed for the prediction of the bioactive compound in HPLC method development, which combines the AI-based models and the classical MLR (i.e FFNN-MLR and SVM-MLR) to enjoy the benefits of the linear and non-linear properties of the models. The results obtained based on the predictive comparison of the single models showed that FFNN outperformed the other two models. Further elucidation of the results showed that the hybrid models FFNN-MLR and SVM-MLR demonstrated higher performance in terms of the performance indices and they are all capable of boosting the performance efficiency of the single models up to 12%.