Ignore:
Timestamp:
11/13/15 18:09:46 (9 years ago)
Author:
sherbold
Message:
  • added CLA/CLAMI data processors (ASE 2015)
  • modified normal WEKA training to have a fallback in case to few instances of a certain class (defect-prone, non-defect-prone) are available: it now uses ZeroR in that case, i.e., a trivial classifier that always predicts the class that appears more often.
File:
1 edited

Legend:

Unmodified
Added
Removed
  • trunk/CrossPare/src/de/ugoe/cs/cpdp/training/WekaBaseTraining.java

    r41 r42  
    5656        classifierParams = Arrays.copyOfRange(params, 2, params.length); 
    5757 
    58         classifier = setupClassifier(); 
     58        //classifier = setupClassifier(); 
    5959    } 
    6060 
     
    6464    } 
    6565 
    66     public Classifier setupClassifier() { 
     66    protected Classifier setupClassifier() { 
    6767        Classifier cl = null; 
    6868        try { 
Note: See TracChangeset for help on using the changeset viewer.