Bytecode and source code of org.tunedit.base.RegressionTT70 class, which implements an evaluation procedure applicable to regression tasks (with continuous decision attribute). The class inherits from org.tunedit.core.EvaluationProcedure base class.
On each run, RegressionTT70 splits the dataset randomly into training and test sets in ratio 70:30. The evaluated algorithm is trained on the training set and then its Root Mean Squared Error (RMSE) is calculated using the test set. Value of RMSE is returned as the result of evaluation.
Important: note that the higher the value of RMSE, the lower the quality of an algorithm. This is opposite to ClassificationTT70, where higher result values correspond to more accurate algorithms.
Always the last attribute of the dataset is taken as the decision one.
Algorithms that can be evaluated by this procedure must be Java classes inheriting either from:
- org.debellor.core.Cell (Debellor), or
- weka.classifiers.Classifier (Weka), or
- rseslib.processing.classification.Classifier (Rseslib).
The dataset must be either a text file in Weka's ARFF format or a Java class that produces the data on request, implemented in Debellor's architecture (Cell subclass).
Classes contained in this JAR: