public static int[,] ComputeConfusionMatrix(this ISupervisedClassifier classifier, IEnumerable <Instance> testingDataset, Feature classFeature)
{
int errors = 0;
var instanceList = testingDataset.ToList();
int[] trueLabel = new int[instanceList.Count];
int[] predicted = new int[instanceList.Count];
for (int i = 0; i < instanceList.Count; i++)
{
Instance instance = testingDataset.ElementAt <Instance>(i);
double[] classification_result = classifier.Classify(instance);
trueLabel[i] = (int)instance[classFeature];
predicted[i] = classification_result.ArgMax <double>();
if (trueLabel[i] != predicted[i])
{
errors++;
}
}
// Computing confusion matrix
int classCount = classFeature.ClassValues().Length;
int[,] confusionMatrix = new int[classCount, classCount];
for (int i = 0; i < trueLabel.Length; i++)
{
confusionMatrix[trueLabel[i], predicted[i]]++;
}
return(confusionMatrix);
}
static void Main(string[] args)
{
Console.WriteLine("Hello World!");
string uciDir = "C:\\Users\\jrenewhite\\OneDrive\\Maestría\\UCI data\\Datasets";
var files = Directory.GetFiles(uciDir).Yaap(settings: new YaapSettings
{
Description = "Reading files",
Width = 100,
ColorScheme = YaapColorScheme.Dark,
SmoothingFactor = 0.5
});
var type = typeof(ISupervisedClassifier);
var classifiersTypes = new List <Type>();
foreach (var assembly in AppDomain.CurrentDomain.GetAssemblies())
{
foreach (Type mytype in assembly.GetTypes()
.Where(mytype => mytype.GetInterfaces().Contains(type)))
{
//do stuff
//Console.WriteLine($"{mytype.Name} is implementable? {mytype.(type)}");
classifiersTypes.Add(mytype);
}
}
List <ISupervisedClassifier> classifiers = new List <ISupervisedClassifier>();
foreach (var classifierType in classifiersTypes)
{
try
{
// test creating an instance of the class.
ISupervisedClassifier classifier = (ISupervisedClassifier)Activator.CreateInstance(classifierType);
classifiers.Add(classifier);
}
catch (Exception e)
{
Console.WriteLine("Could not create {0} class.", classifierType);
}
}
foreach (var item in files)
{
try
{
var serializer = new ARFFSerializer(item);
Dataset dataset = serializer.DeserializeDataset();
VIC vic = new VIC(dataset, classifiers, new AccordKMeans());
double index = vic.GetValidityIndex();
}
catch (Exception e)
{
Console.WriteLine(e.Message);
}
}
}
public static Tuple <double, double, double, double, double> TestClassifier(this ISupervisedClassifier classifier, IEnumerable <Instance> testingDataset, Feature classFeature)
{
int errors = 0;
var instanceList = testingDataset.ToList();
int[] trueLabel = new int[instanceList.Count];
int[] predicted = new int[instanceList.Count];
for (int i = 0; i < instanceList.Count; i++)
{
Instance instance = testingDataset.ElementAt <Instance>(i);
double[] classification_result = classifier.Classify(instance);
trueLabel[i] = (int)instance[classFeature];
predicted[i] = classification_result.ArgMax <double>();
if (trueLabel[i] != predicted[i])
{
errors++;
}
}
// Computing confusion matrix
int classCount = classFeature.ClassValues().Length;
int[,] confusionMatrix = new int[classCount, classCount];
for (int i = 0; i < trueLabel.Length; i++)
{
confusionMatrix[trueLabel[i], predicted[i]]++;
}
double acc = 100.0 * (instanceList.Count - errors) / instanceList.Count;
double auc = ComputeMultiClassAUC(confusionMatrix);
double precision = ComputeMultiClassPrecision(confusionMatrix);
double recall = ComputeMultiClassRecall(confusionMatrix);
double f1 = ComputeMultiClassF1(confusionMatrix);
return(new Tuple <double, double, double, double, double>(acc, auc, precision, recall, f1));
}
