For comparison with previous methods, we chose a "one-to-the-rest" method using a support vector machine from the first type [ 14], and the shrunken centroid method [ 18] from the latter type.
When compared to the rest methods, tuned DNNs were found to statistically outperform, with p value <0.01 based on Wilcoxon statistical test.
DNN overall outperformed the rest methods across most tested datasets, when dealt with noisy data their performance in several datasets dropped below that of other methods included in the study indicating that these methods might be more sensitive to noise that the rest methods.
Wilcoxon paired signed-rank statistical test was applied, with confidence intervals 99%, to compare the performance achieved by DNN against each of the rest methods, results are shown in Table 5.
Instead, at higher level of noise equal or higher than 30% Naïve Bayes reached or even outperformed the rest methods, e.g. in cases of datasets CDK2 and CB1, demonstrating higher tolerance to noise.
However, V-ELM is better than the rest methods in general when using the nonlinear kernels.
In order to determine which features are relevant for classification, we performed a feature selection analysis in which one signal is chosen as a proxy for a classification value and is compared against the rest (Methods).
"one-to-the-rest" methods [ 9, 10], pairwise comparisons, i.e., "one-to-one" methods [ 11- 13], error-correcting output coding [ 21- 23] belong to this type.
The relative expression software tool (REST) method was used for statistical analysis of real-time PCR data (Pfaffl et al., 2002).
The rest cure method was designed by a neurologist, Silas Weir Mitchell, in the late 19th century in the United States.
The SC-FIS method has better performance than the rest three methods in the prediction of each of the soil microbial dynamics (R2 of 0.99 in the prediction of PS).
