public IBClassificationQualityEvaluator(AbstractLazyClassifier classifer, IClassificationMeasure measure)
: base(0, 1)
{
this._IBClassifier = classifer;
this._measure = measure;
//this.LoadSimilarityCache();
}
public DataMining.Model.IClassifier CreateClassifier()
{
this.Work();
this._problem.SolutionQualityEvaluator.EvaluateSolutionQuality(this.GlobalBest);
AbstractLazyClassifier clasifier = ((IBClassificationQualityEvaluator)this._problem.SolutionQualityEvaluator).Classifier;
return(clasifier);
}
public DataMining.Model.IClassifier CreateClassifier()
{
this.Initialize();
this.Evolve();
this._solutionQualityEvaluator.EvaluateSolutionQuality(this.GlobalBest);
AbstractLazyClassifier clasifier = ((IBClassificationQualityEvaluator)this._solutionQualityEvaluator).Classifier;
return(clasifier);
}
public DataMining.Model.IClassifier CreateClassifier()
{
this.Work();
double[] weights = this.GlobalBest.ToList().ToArray();
AbstractLazyClassifier clasifier = ((IBClassificationQualityEvaluator)this._problem.SolutionQualityEvaluator).Classifier;
clasifier.Weights = weights;
return(clasifier);
}
public IBLHeuristicsCalculator(Dataset dataset, AbstractLazyClassifier lazyClassifier)
{
this._dataset = dataset;
this._lazyClassifier = lazyClassifier;
this._probabilities = new Dictionary <int, double>();
for (int i = 0; i < this._dataset.Size; i++)
{
Prediction predicition = this._lazyClassifier.Classify(this._dataset[i]);
this._probabilities.Add(i, predicition.Probabilities[predicition.Label]);
}
}
private static void SaveWeights(Dataset dataset, AbstractLazyClassifier lazyClassifier)
{
System.IO.StreamWriter writer = new System.IO.StreamWriter("results\\" + dataset.Metadata.DatasetName + ".txt", true);
string header = "att\\class,";
foreach (string classlabel in dataset.Metadata.Target.Values)
{
header += classlabel + ",";
}
header = header.Substring(0, header.Length - 1);
writer.WriteLine(header);
for (int classIndex = 0; classIndex < dataset.Metadata.Target.Values.Length; classIndex++)
{
double sum = 0;
for (int attributeIndex = 0; attributeIndex < dataset.Metadata.Attributes.Length; attributeIndex++)
{
sum += lazyClassifier.Weights[classIndex][attributeIndex];
}
for (int attributeIndex = 0; attributeIndex < dataset.Metadata.Attributes.Length; attributeIndex++)
{
lazyClassifier.Weights[classIndex][attributeIndex] /= sum;
}
}
for (int attributeIndex = 0; attributeIndex < dataset.Metadata.Attributes.Length; attributeIndex++)
{
string line = dataset.Metadata.Attributes[attributeIndex].Name + ",";
double[] weights = new double[dataset.Metadata.Target.Values.Length];
for (int classIndex = 0; classIndex < dataset.Metadata.Target.Values.Length; classIndex++)
{
weights[classIndex] = Math.Round((lazyClassifier.Weights[classIndex][attributeIndex]) * 100, 1);
}
line += string.Join(",", weights);
writer.WriteLine(line);
}
writer.Flush();
writer.Close();
}
public DataMining.Model.Ensembles.EnsembleClassifier CreateEnsembleClassifier()
{
this.Work();
List <DataMining.Model.ClassifierInfo> ensemble = new List <DataMining.Model.ClassifierInfo>();
for (int i = 0; i < this._archiveSize; i++)
{
Solution <double> solution = Solution <double> .FromList(this._solutionArchive[i].ToList());
this._problem.SolutionQualityEvaluator.EvaluateSolutionQuality(solution);
AbstractLazyClassifier clasifier = ((IBClassificationQualityEvaluator)this._problem.SolutionQualityEvaluator).Classifier;
DataMining.Model.ClassifierInfo classifierInfo = new DataMining.Model.ClassifierInfo()
{
Desc = clasifier.Desc + "-" + i.ToString(), Classifier = clasifier, Quality = solution.Quality
};
ensemble.Add(classifierInfo);
}
DataMining.Model.Ensembles.EnsembleClassifier ensmebleClassifier = new DataMining.Model.Ensembles.EnsembleClassifier(ensemble);
return(ensmebleClassifier);
}
