Where, λ i is the matrix eigenvalue, and the is the Hermite matrix.
2.4 Model Validation
These QSPR models were quantitatively assessed by several statistical criteria as relative deviation (RD ), the Fisher significance parameter (F ) and the average absolute relative deviation percentage (AARD %). The predictability of the model is supported by R2training for the training set and R2testing for the testing set. In addition, Leave-one-out (LOO) validation was utilized to evaluate the robustness of this QSPR model (Q 2). Y -randomization test was utilized to avoid the possibility of chance correlation in the modelling work.32,49