Ia fix 377 core #383
Ia fix 377 core #383
Conversation
idhamari
commented
May 3, 2022
idhamari
commented
- Added method to evaluate with a testing subset
- Added a new example to show the use case
- This modification will be used in the next commit for modifying the GUI so user can select to evaluate with K-fold or not
| @@ -74,7 +74,8 @@ public static ClassificationConfigurationRandomForest createRandomForest() { | |||
| private int vectorLength = DEFAULT_VECTOR_LENGTH; | |||
| /** Modified */ | |||
| private boolean modified = false; | |||
|
|
|||
| /** EvaluateWithKfold */ | |||
| private boolean EvaluateWithKfold = true; | |||
There was a problem hiding this comment.
Java properties start with lower case letters.
| * This class implements an example on how to compare data mining performance | ||
| * It shows how to use subset for training and different subset for testing | ||
| * @author Fabian Prasser | ||
| * @author Florian Kohlmayer |
| * @author Fabian Prasser | ||
| * @author Florian Kohlmayer | ||
| */ | ||
| public class Example39_subset extends Example { |
There was a problem hiding this comment.
Convention for examples is Example[NUMBER]
| data.getDefinition().setResponseVariable("marital-status", true); | ||
|
|
||
| // Create a training subset data with a specific percentage of the original data e.g 80% | ||
|
|
There was a problem hiding this comment.
File needs to be cleaned up. Remove blank lines, unnecessary comments etc.
| @@ -74,7 +74,8 @@ public static ClassificationConfigurationRandomForest createRandomForest() { | |||
| private int vectorLength = DEFAULT_VECTOR_LENGTH; | |||
| /** Modified */ | |||
| private boolean modified = false; | |||
|
|
|||
| /** EvaluateWithKfold */ | |||
There was a problem hiding this comment.
Choose nice description, e.g. "Use k-fold cross validation"
| // For each fold as a validation set | ||
| for (int evaluationFold = 0; evaluationFold < folds.size(); evaluationFold++) { | ||
|
|
||
| if (config.getEvaluateWithKfold()) { |
There was a problem hiding this comment.
The two versions should be put into two different methods, e.g.
evaluateWithCrossValidation(...)
evaluateWithTrainingSet(...)
Currently it is (a) hard to review, (b) there are some redundancies. For example, you are calculating the folds in line 351, even if you dont need them because you are using a training and test set.
| @@ -450,7 +547,7 @@ private static ClassificationMethod getClassifier(WrappedBoolean interrupt, | |||
| } else { | |||
| throw new UnexpectedErrorException(e); | |||
| } | |||
| } | |||
| } | |||
|
Also please refer to "#377" in your description so that this gets linked with the issue. |
|
|
Merged other PR into feature branch. Closing. |
