static void Main(string[] args)
{
// read the data
string line;
StreamReader dataFile = new StreamReader("E:\\documents\\com\\pasca\\patternrec\\RBF_PNN\\wdbc-normal.data");
List <Vector> data = new List <Vector>();
List <int> classLabel = new List <int>();
while ((line = dataFile.ReadLine()) != null)
{
if (line.IndexOf('?') == -1)
{
string[] dataLine = line.Split(',');
double[] dataLineNum = new double[30];
for (int i = 0; i < 30; i++)
{
dataLineNum[i] = double.Parse(dataLine[i + 2]);
}
data.Add(new Vector(dataLineNum));
// 1 = Benign, 2 = Malignant
int c = (dataLine[1] == "B") ? 1 : 2;
classLabel.Add(c);
}
}
dataFile.Close();
// K-means for centroid calculation
int k = 2;
KMeans kmeans = new KMeans(k);
for (int i = 0; i < data.Count; i++)
{
kmeans.AddData(data[i]);
}
kmeans.DoClustering(100, 0.001);
// build the RBF
RBF rbf = new RBF(data.Count, 30, k, 1);
for (int i = 0; i < data.Count; i++)
{
rbf.AddTrainingData(classLabel[i], data[i]);
}
// center calculated using K-means
Matrix rbfCenter = new Matrix(k, 30);
for (int i = 0; i < k; i++)
{
rbfCenter.SetRow(i, kmeans.GetCentroid()[i].GetValueArray());
}
rbf.SetCenter(rbfCenter);
rbf.Train();
// testing
List <int> rbfActualClass = classLabel;
List <int> rbfPredictedClass = new List <int>();
for (int i = 0; i < data.Count; i++)
{
rbfPredictedClass.Add(rbf.Simulate(data[i]));
}
Console.WriteLine("RBF:\n");
ConfusionMatrix rbfConfusionMat = new ConfusionMatrix(rbfPredictedClass, rbfActualClass);
rbfConfusionMat.Compute();
rbfConfusionMat.Print();
Console.WriteLine("Accuracy: " + rbfConfusionMat.GetAccuracy());
// build the PNN
PNN pnn = new PNN(2, new int[] { 1, 2 });
for (int i = 0; i < data.Count; i++)
{
pnn.AddTrainingData(classLabel[i], data[i]);
}
pnn.Train();
// testing
List <int> pnnActualClass = classLabel;
List <int> pnnPredictedClass = new List <int>();
for (int i = 0; i < data.Count; i++)
{
pnnPredictedClass.Add(pnn.Simulate(data[i], 1));
}
Console.WriteLine("\n\nPNN:\n");
ConfusionMatrix pnnConfusionMat = new ConfusionMatrix(pnnPredictedClass, pnnActualClass);
pnnConfusionMat.Compute();
pnnConfusionMat.Print();
Console.WriteLine("Accuracy: " + pnnConfusionMat.GetAccuracy());
Console.Read();
}
static void Main(string[] args)
{
// read the data
string line;
StreamReader dataFile = new StreamReader("E:\\documents\\com\\pasca\\patternrec\\RBF_PNN\\wdbc-normal.data");
List<Vector> data = new List<Vector>();
List<int> classLabel = new List<int>();
while ((line = dataFile.ReadLine()) != null)
{
if (line.IndexOf('?') == -1)
{
string[] dataLine = line.Split(',');
double[] dataLineNum = new double[30];
for (int i = 0; i < 30; i++) dataLineNum[i] = double.Parse(dataLine[i + 2]);
data.Add(new Vector(dataLineNum));
// 1 = Benign, 2 = Malignant
int c = (dataLine[1] == "B") ? 1 : 2;
classLabel.Add(c);
}
}
dataFile.Close();
// K-means for centroid calculation
int k = 2;
KMeans kmeans = new KMeans(k);
for (int i = 0; i < data.Count; i++)
{
kmeans.AddData(data[i]);
}
kmeans.DoClustering(100, 0.001);
// build the RBF
RBF rbf = new RBF(data.Count, 30, k, 1);
for (int i = 0; i < data.Count; i++)
{
rbf.AddTrainingData(classLabel[i], data[i]);
}
// center calculated using K-means
Matrix rbfCenter = new Matrix(k, 30);
for (int i = 0; i < k; i++)
{
rbfCenter.SetRow(i, kmeans.GetCentroid()[i].GetValueArray());
}
rbf.SetCenter(rbfCenter);
rbf.Train();
// testing
List<int> rbfActualClass = classLabel;
List<int> rbfPredictedClass = new List<int>();
for (int i = 0; i < data.Count; i++)
{
rbfPredictedClass.Add(rbf.Simulate(data[i]));
}
Console.WriteLine("RBF:\n");
ConfusionMatrix rbfConfusionMat = new ConfusionMatrix(rbfPredictedClass, rbfActualClass);
rbfConfusionMat.Compute();
rbfConfusionMat.Print();
Console.WriteLine("Accuracy: " + rbfConfusionMat.GetAccuracy());
// build the PNN
PNN pnn = new PNN(2, new int[] { 1, 2 });
for (int i = 0; i < data.Count; i++)
{
pnn.AddTrainingData(classLabel[i], data[i]);
}
pnn.Train();
// testing
List<int> pnnActualClass = classLabel;
List<int> pnnPredictedClass = new List<int>();
for (int i = 0; i < data.Count; i++)
{
pnnPredictedClass.Add(pnn.Simulate(data[i], 1));
}
Console.WriteLine("\n\nPNN:\n");
ConfusionMatrix pnnConfusionMat = new ConfusionMatrix(pnnPredictedClass, pnnActualClass);
pnnConfusionMat.Compute();
pnnConfusionMat.Print();
Console.WriteLine("Accuracy: " + pnnConfusionMat.GetAccuracy());
Console.Read();
}