public static void decisionTreeDemo()
{
try
{
DataSet ds = DataSetFactory.getRestaurantDataSet();
DecisionTreeLearner learner = new DecisionTreeLearner();
learner.Train(ds);
System.Console.WriteLine("The Induced Decision Tree is ");
System.Console.WriteLine(learner.getDecisionTree());
int[] result = learner.Test(ds);
System.Console.WriteLine("\nThis Decision Tree classifies the data set with "
+ result[0]
+ " successes"
+ " and "
+ result[1]
+ " failures");
System.Console.WriteLine("\n");
}
catch (Exception e)
{
System.Console.WriteLine("Decision Tree Demo Failed ");
throw e;
}
}
public void testInducedDecisionTreeClassifiesRestaurantDataSetCorrectly()
{
DecisionTreeLearner learner = new DecisionTreeLearner(
createInducedRestaurantDecisionTree(), "Unable to clasify");
int[] results = learner.test(DataSetFactory.getRestaurantDataSet());
Assert.AreEqual(12, results[0]);
Assert.AreEqual(0, results[1]);
}
public void testInducedTreeClassifiesDataSetCorrectly()
{
DataSet ds = DataSetFactory.getRestaurantDataSet();
DecisionTreeLearner learner = new DecisionTreeLearner();
learner.train(ds);
int[] result = learner.test(ds);
Assert.AreEqual(12, result[0]);
Assert.AreEqual(0, result[1]);
}
public void testDefaultUsedWhenTrainingDataSetHasNoExamples()
{
// tests RecursionBaseCase#1
DataSet ds = DataSetFactory.getRestaurantDataSet();
DecisionTreeLearner learner = new DecisionTreeLearner();
DataSet ds2 = ds.emptyDataSet();
Assert.AreEqual(0, ds2.size());
learner.train(ds2);
Assert.AreEqual("Unable To Classify", learner.predict(ds
.getExample(0)));
}
public void testStumpPredictionForDataSet()
{
DataSet ds = DataSetFactory.getRestaurantDataSet();
List <String> unmatchedValues = new List <String>();
unmatchedValues.Add(NO);
DecisionTree tree = DecisionTree.getStumpFor(ds, "hungry", YES, YES,
unmatchedValues, "Unable to Classify");
DecisionTreeLearner learner = new DecisionTreeLearner(tree,
"Unable to Classify");
int[] result = learner.test(ds);
Assert.AreEqual(5, result[0]);
Assert.AreEqual(7, result[1]);
}
public void testClassificationReturnedWhenAllExamplesHaveTheSameClassification()
{
// tests RecursionBaseCase#2
DataSet ds = DataSetFactory.getRestaurantDataSet();
DecisionTreeLearner learner = new DecisionTreeLearner();
DataSet ds2 = ds.emptyDataSet();
// all 3 examples have the same classification (willWait = yes)
ds2.add(ds.getExample(0));
ds2.add(ds.getExample(2));
ds2.add(ds.getExample(3));
learner.train(ds2);
Assert.AreEqual("Yes", learner.predict(ds.getExample(0)));
}
public void BestFirstTreeBuilder_Build_Leaf_Nodes_4()
{
var(observations, targets) = DataSetUtilities.LoadGlassDataSet();
var sut = new DecisionTreeLearner(new BestFirstTreeBuilder(2000, 4, observations.ColumnCount, 0.000001, 42,
new OnlyUniqueThresholdsSplitSearcher(1), new GiniClassificationImpurityCalculator()));
var model = new ClassificationDecisionTreeModel(sut.Learn(observations, targets));
var predictions = model.Predict(observations);
var evaluator = new TotalErrorClassificationMetric <double>();
var actual = evaluator.Error(targets, predictions);
Assert.AreEqual(0.37383177570093457, actual, 0.00001);
}
public void testMajorityReturnedWhenAttributesToExamineIsEmpty()
{
// tests RecursionBaseCase#2
DataSet ds = DataSetFactory.getRestaurantDataSet();
DecisionTreeLearner learner = new DecisionTreeLearner();
DataSet ds2 = ds.emptyDataSet();
// 3 examples have classification = "yes" and one ,"no"
ds2.add(ds.getExample(0));
ds2.add(ds.getExample(1));// "no"
ds2.add(ds.getExample(2));
ds2.add(ds.getExample(3));
ds2.setSpecification(new MockDataSetSpecification("will_wait"));
learner.train(ds2);
Assert.AreEqual("Yes", learner.predict(ds.getExample(1)));
}
public void BestFirstTreeBuilder_Build_Leaf_Nodes_4()
{
var parser = new CsvParser(() => new StringReader(Resources.Glass));
var observations = parser.EnumerateRows(v => v != "Target").ToF64Matrix();
var targets = parser.EnumerateRows("Target").ToF64Vector();
var rows = targets.Length;
var sut = new DecisionTreeLearner(new BestFirstTreeBuilder(2000, 4, observations.ColumnCount, 0.000001, 42,
new OnlyUniqueThresholdsSplitSearcher(1), new GiniClassificationImpurityCalculator()));
var model = new ClassificationDecisionTreeModel(sut.Learn(observations, targets));
var predictions = model.Predict(observations);
var evaluator = new TotalErrorClassificationMetric <double>();
var actual = evaluator.Error(targets, predictions);
Assert.AreEqual(0.37383177570093457, actual, 0.00001);
}
ClassificationDecisionTreeModel CreateTree(F64Matrix observations, double[] targets, int[] indices, Random random)
{
var learner = new DecisionTreeLearner(
new DepthFirstTreeBuilder(m_maximumTreeDepth,
m_featuresPrSplit,
m_minimumInformationGain,
m_random.Next(),
new RandomSplitSearcher(m_minimumSplitSize, m_random.Next()),
new GiniClasificationImpurityCalculator()));
var treeIndicesLength = (int)Math.Round(m_subSampleRatio * (double)indices.Length);
var treeIndices = new int[treeIndicesLength];
for (int j = 0; j < treeIndicesLength; j++)
{
treeIndices[j] = indices[random.Next(indices.Length)];
}
var model = new ClassificationDecisionTreeModel(learner.Learn(observations, targets, treeIndices));
return(model);
}