In the case of training data set, the coefficient of determination (R) and absolute average deviation (AAD) were 1 and 0.002844, respectively, whereas for the testing data set, R2 was 0.994122 and AAD was 1.289405 (Table 4) and for validating data sets R2 and AAD were, 0.99998696 and 1.377, respectively (Table 5).
In the case of training data set, the coefficient of determination (R) and absolute average deviation (AAD) were 1 and 0.1%, respectively, whereas for the testing data set, R2 was 1 and AAD was 0.231% (Table 4) and for validating data sets R2 and AAD were 0.989 and 0.059%, respectively (Table 5).
The selection of training data is critical.
Mean absolute percent error (MAPE) of 1.81% and coefficient of determination (R2) of 0.9976 for training data and MAPE of 1.52% and R2 value of 0.9948 for testing data were obtained.
Results in Figure 3 show that accuracy increased more with the size of the training data for traits with a low estimate of π (interpreted as determination by many loci) and high heritability.
The best statistical coefficient of multiple determinations (R2-value) for training data equals to 0.999715, 0.995627, 0.999497, and 0.997648 obtained by different algorithms with seven neurons for the non-dimensional exergy losses of evaporator, generator, absorber and condenser, respectively.
Fig. 5 Influence of extended training data.
The solution consists of adding another input to the standard feed-forward network and of modifying the training algorithm to allow for determination of this input for which no training data is available.
Specifically, SVM-based methods calculate separating hyperplanes, which are determined by only some training data, i.e., support vectors, and all other training data are irrelevant to the determination of hyperplanes.
Coefficient of determination (R 2) of 0.99985 and 0.99505 are acquired for training data set of As III) and As V), respectively.
