Abstract
This research aimed at modelling and predicting the antioxidant activities of Amaranthus viridis seed extract using four (4) data-driven models. Artificial Neural Network (ANN), Support Vector Machine (SVM), k-nearest Neighbour (k-NN), and Decision Tree (DT) were used as modelling algorithms for the construction of a non-linear empirical model to predict the antioxidant properties of Amaranthus viridis seed extract. Datasets for the modelling operation were obtained from a Box Behnken design while the hyperparameters of the ANN, SVM, k-NN and DT were determined using a 10-fold cross-validation technique. Among the Machine Learning algorithms, DT was observed to exhibit excellent performance and outperformed other Machine Learning algorithms in predicting the antioxidant activities of the seed extract, with a sensitivity of 0.867, precision of 0.928, area under the curve of 0.979, root mean square error of 0.184 and correlation coefficient of 0.9878. It was closely followed by ANN which was used to analyze and explain in detail the effect of the independent variables on the antioxidant activities of the seed extracts. This result affirmed the suitability of DT in predicting the antioxidant activities of Amaranthus viridis.