public void DecisionTreeLearn_ThreeSplit()
{
GlobalVariables.GrowCondition = new GrowCondition(7);
var synthesisTree = new SynthesisTreeLearn(attributes, learn, validation);
synthesisTree.Learn();
var tree = synthesisTree.Tree;
Assert.AreEqual(1, tree.Root.Branches.Count);
Assert.AreEqual(1, tree.Root.Branches[0].Branches.Count);
Assert.AreEqual(1, tree.Root.Branches[0].Branches[0].Branches.Count);
Assert.AreEqual(1, tree.Root.Branches[0].Branches[0].Branches[0].Branches.Count);
Assert.AreEqual(3, tree.Root.Branches[0].Branches[0].Branches[0].Branches[0].Branches[0].Branches[0].Branches[0].Output);
Assert.AreEqual(0, tree.Root.Branches[0].Branches[0].Branches[0].Branches[0].Branches[0].Branches[0].Branches[1].Output);
Assert.AreEqual(9, tree.Root.Branches[0].Branches[0].Branches[0].Branches[0].Branches[0].Branches[1].Output);
Assert.AreEqual(1, tree.Root.Branches[0].Branches[0].Branches[0].Branches[0].Branches[1].Branches[0].Output);
Assert.AreEqual(2, tree.Root.Branches[0].Branches[0].Branches[0].Branches[0].Branches[1].Branches[1].Output);
int[] actual = tree.Decide(validation)
.Select(val => val > 0 ? 0 : 1)
.ToArray();
var areaSize = (double)tree.Decide(validation)
.Count(val => val > 0) / validation.Length;
var confusionMatrix = new Accord.Statistics.Analysis.ConfusionMatrix(actual, validationOutput, 0, 1);
}
private static void CalculateStatistics(ExperimentSettings settings)
{
using (var database = new Database(settings.DatabaseFilePath))
{
using (var experiment = database.NewExperiment())
{
try
{
experiment.SaveIn(settings);
experiment["out_totalTime"] = timer.ElapsedMilliseconds;
experiment["out_decisionTreeHeight"] = decisionTree.GetHeight();
experiment["in_terms"] = dataStorage.TestInputDataSet[0].Length;
if (decisionTree == null)
{
throw new ArgumentNullException(nameof(decisionTree));
}
int[] actual = decisionTree.Decide(dataStorage.TestInputDataSet)
.Select(val => val > 0 ? 0 : 1)
.ToArray();
var confusionMatrix = new Accord.Statistics.Analysis.ConfusionMatrix(actual, dataStorage.TestOutputDataSet, 0, 1);
var validationPropablity =
(double)decisionTree.Decide(dataStorage.ValidationDataSet)
.Count(val => val > 0) / dataStorage.ValidationDataSet.Length;
experiment["in_test_examples"] = dataStorage.TestOutputDataSet.Length;
experiment["in_test_positives"] = confusionMatrix.ActualPositives;
experiment["in_test_negatives"] = confusionMatrix.ActualNegatives;
experiment["in_learn_examples"] = dataStorage.LearnDataSet.Length;
experiment["in_validation_examples"] = dataStorage.ValidationDataSet.Length;
experiment["out_accurancy"] = confusionMatrix.Accuracy;
experiment["out_error"] = confusionMatrix.Error;
experiment["out_sensitivity"] = confusionMatrix.Sensitivity;
experiment["out_specificity"] = confusionMatrix.Specificity;
experiment["out_efficiency"] = confusionMatrix.Efficiency;
experiment["out_precision"] = confusionMatrix.Precision;
experiment["out_recall"] = confusionMatrix.Recall;
experiment["out_fScore"] = confusionMatrix.FScore;
experiment["out_jaccardIndex"] =
(double)confusionMatrix.TruePositives / (confusionMatrix.TruePositives +
confusionMatrix.FalseNegatives +
confusionMatrix.FalsePositives);
experiment["out_falsePositives"] = confusionMatrix.FalsePositives;
experiment["out_falseNegatives"] = confusionMatrix.FalseNegatives;
experiment["out_trueNegatives"] = confusionMatrix.TrueNegatives;
experiment["out_truePositives"] = confusionMatrix.TruePositives;
experiment["out_measure"] =
StatisticsCalculator.CalculateObjectiveFunction(confusionMatrix.Recall, validationPropablity);
experiment["out_matthewsCorrelationCoefficient"] = confusionMatrix.MatthewsCorrelationCoefficient;
var rules = decisionTree.ToRules();
var rules2 = new List <Rule>();
foreach (var rule in rules)
{
rules2.Add(new Rule(rule.ToString().Replace(",", ".")));
}
var constraints = rules2.ReturnUniqueSplitsWithBooleans();
experiment["out_constraints"] = constraints.Item1.Count;
}
catch (Exception e)
{
experiment["error"] = e.Message;
}
finally
{
experiment.Save();
}
}
}
}
