Response surface methodology (RSM), which includes experimental design, model fitting, validation and condition optimization, has eliminated the drawbacks of single-factor experimental design and been proved to be powerful and useful for the optimization of ATPS [14], [15].
Model fitting, validation, adaptation for studies of genetic variability, as well as analysis of model predictions and interpretation of results was done with support of the RSCF grant no.
In the training phase, different ensembles of classifiers are adaptively generated by fitting the validation data globally with different degrees.
Our research groups use a variety of the many features that Origin offers including linear and non-linear curve fitting, model validation, dataset comparison tools, and multi-dimensional data analysis.
Creation of an optimal predictive model demands many critical and time-consuming steps, including data collection and processing, appropriate data representation (descriptors and fingerprints calculation), evaluation of the data set modelability, best predictors selection, machine learning models fitting and validation.
A shrinkage factor was applied to log odds ratios after model fitting before validation [ 29].
Furthermore, many existing classification methods require predetermination of a set of predictor variables, thereby introducing additional complexity and bias that could adversely affect both model fitting and validation (Ambroise and McLachlan 2002).
The outputs of this program included the descriptive statistics (Figs. 14, 15), parameter estimation, statistical hypothesis testing, and the basic linear regression which consists of linear regression, analysis of variance (ANOVA) table, and model fit validation (Figs. 16, 17).
Since the equations for response parameters have been derived from quadratic regression fits, therefore validation test should be performed for verification of the legitimacy.
The final decision is made by taking into consideration both the ability of each ensemble to fit the validation data locally and reducing the risk of overfitting.
We found GAM's to perform as well as nonlinear models with respect to fit and validation statistics; however, when applied to the population of predictor data the bootstrap confidence interval was less efficient than the null model.
