Relative differences in moments E (f (X ) ) for data split at X = 250 and X = 350 for the Schlögl system as percentages of their values when split at X = 300, i.e. 100 ∗ | E (f (X ) ) split 350 − E (f (X ) ) split 250 | / E (f (X ) ) split 300.
Particularly, for each data splitting and CV method, the following plots are produced: observed versus predicted values for training/test sets, feature importance, fitted versus residuals for the fitted model, leverage statistics for fitted model, Cook's distance for fitted model, and six standard fitting plots including Cook's distance cutoff.
A CSV output file is created with all the basic statistics (17 values) for each method, data splitting and CV type.
It is also advisable to study the variable selection frequencies for different data splits and test data sizes.
The map tasks are created only for new data splits while the reduce tasks take their input from both the current map output and the map output of the previous computation.
Sample sizes for clicker data, split by treatment and instructor, are shown in Table 1.
However, because of the data splitting necessary for cross-validation, the number of data instances in one part of the data would have been too small for the statistical test we used to select descriptors.
Further work has focused on alternative methods for selecting weights, including a data splitting technique proposed by Roeder and Wasserman [ 9].
In general, Co-training requires two independent and sufficient [6] feature sets (called views), yet for single view data sets, view splitting and training data splitting can be used for generating two classifiers, so as to achieve the performance improvements with co-training style learners.
This data splitting was performed for 100 iterations followed by computing the mean values for the performance metrics.
