The percentage average absolute deviation (%AAD) was employed as the criterion for comparing the predictive capabilities of the aforementioned models.
The percentage average absolute deviation (AAD) of the property prediction model that this study used to estimate the value of heat of formation is 4.67%.
The effects of the training algorithm, neural network architectures and transfer function on the ANN performance, as reflected by the percentage average absolute deviation are discussed.
Different calibration procedures have been incorporated and model behaviour is evaluated based on the percentage of absolute deviation (Kanellopoulos and others 2010).
Comparisons indicate the percentage of absolute deviations in PρT, saturated liquid density, and compressibility factors which are lower than 0.5%, 1%, and 2.5%, respectively.
Two criteria, the regression coefficient (R2) and the percentage Absolute Error of Deviation (AED) between experimental and calculated results were used to evaluate the validity of the models.
Linear, quadratic, exponential and S-Curve models were sequentially fitted to the data and for each model its adequacy was assessed using the mean absolute percentage error, mean absolute deviation and the mean squared deviation.
The average absolute deviation pressure percentage (AADP%) for most of the cases studied is observed to be well within 10%, thus proving its efficacy.
We also calculated percentage of scores with an absolute deviation of one point or less from the median score of raters 1 to 7.
In order to compare the goodness of fit between the data calculated from the simulation software and the data acquired by the logs of the installation, we used the following statistical parameters: root mean square error (RMSE), mean absolute deviation (MAD), absolute percentage error (MAPE) and model efficiency (EF).
