// Similarity comparison
public double CompareTo(BreastCancerCase other, List <float> meanWeights,
List <float> errorWeights, List <float> extremeWeights)
{
List <double> values = new List <double>();
values.Add(this.Mean.CalculateSimilarity(other.Mean, meanWeights));
values.Add(this.StandardError.CalculateSimilarity(other.StandardError, errorWeights));
values.Add(this.Extreme.CalculateSimilarity(other.Extreme, extremeWeights));
return(values.Average());
}
// Creates a new randomized experiment setup, splitting the data in two parts
public ExperimentSetup(List <BreastCancerCase> dataset, double testRatio)
{
int testCount = (int)Math.Round(dataset.Count * testRatio);
Random rng = new Random();
int n = dataset.Count;
while (n > 1)
{
n--;
int k = rng.Next(n + 1);
BreastCancerCase value = dataset[k];
dataset[k] = dataset[n];
dataset[n] = value;
}
CaseBase = dataset.GetRange(0, dataset.Count - testCount);
TestSet = dataset.GetRange(dataset.Count - testCount, testCount);
TestRatio = testRatio;
}
static double RunExperiment(ExperimentSetup setup, bool print = false)
{
int n = setup.TestSet.Count;
int[,] confusionMatrix = new int[n, n]; // Implicitly filled with zeros
foreach (BreastCancerCase testCase in setup.TestSet)
{
List <BreastCancerCase> ordered = setup.CaseBase
.OrderByDescending(value => value.CompareTo(
testCase, setup.MeanWeights, setup.ErrorWeights, setup.ExtremeWeights
)).ToList();
if (print)
{
Console.WriteLine("Case under evaluation: ");
testCase.Print();
Console.WriteLine("\nMost similar cases in the database:\n");
}
bool actual = testCase.Malignant;
bool predicted = ordered[0].Malignant;
int row = predicted ? 1 : 0;
int column = actual ? 1 : 0;
confusionMatrix[row, column]++;
if (print)
{
for (int i = 0; i < 5; i++)
{
BreastCancerCase match = ordered[i];
double y = match.CompareTo(testCase, setup.MeanWeights, setup.ErrorWeights, setup.ExtremeWeights);
Console.WriteLine("Match n. #" + i);
match.Print();
bool valid = match.Malignant == actual;
Console.WriteLine("\nSimilarity: " + y);
Console.WriteLine("Valid diagnosis: " + (valid ? "YES" : "NO"));
Console.WriteLine();
}
}
}
// True Negatives + True Positives / Negatives + Positives
double accuracy = (double)(confusionMatrix[0, 0] + confusionMatrix[1, 1]) / (double)n;
if (print)
{
Console.WriteLine("Outcome:");
Console.WriteLine("\tB\tM");
Console.WriteLine("B\t" + confusionMatrix[0, 0] + "\t" + confusionMatrix[0, 1]);
Console.WriteLine("M\t" + confusionMatrix[1, 0] + "\t" + confusionMatrix[1, 1]);
Console.WriteLine("Accuracy: " + accuracy);
}
return(accuracy);
}