2.3 Model validation and evaluation
The statistical parameters, including squared correlation coefficient
(R2 ), average absolute error (AAE), and
correlation coefficient for leave-one-out cross validation
(Q 2LOO-CV) are employed to
evaluate the goodness-of-fit of the model. The definitions of these
statistical parameters are given in Eqs (2)-(4). Additionally, the
testing set is used in the external validation to assess the external
predictive abilities of the model. To evaluate the model’s robustness,
the internal validation results are offered. Application domain analysis
and Y-randomized analysis are performed to confirm the reliability of
the model and to exclude chance-related aspects from the modeling
process.
Note: T g i,exp and T g
i,cal represent experimental and calculated values of the polyesters,
respectively. n is the number of data points.n train is the number of data points in the
training set.