/// <summary>
/// Runs the ELM network and stores the predictions
/// </summary>
/// <param name="sender"></param>
/// <param name="e"></param>
private void button_Predict_Click(object sender, EventArgs e)
{
prediction = elm.predict(elm.Xtest);
test = elm.Ytest;
output_SampleCount.Text = (prediction.RowCount + 1).ToString();
input_SelectedSample.Enabled = true;
button_Plot.Enabled = true;
button_PlotBestSample.Enabled = true;
}
/// <summary>
/// Initializes the statistics window with the appropriate testing type
/// </summary>
/// <param name="type"></param>
/// <param name="testOffset"></param>
private void ShowStats(String type, int testOffset)
{
int inputDays = Int32.Parse(input_InputDays.Text);
int outputDays = Int32.Parse(input_OutputDays.Text);
int hiddenLayerSize = Int32.Parse(input_HiddenLayerSize.Text);
int trainPercentage = Int32.Parse(input_TrainPercentage.Text);
int biasValue = Int32.Parse(input_BiasValue.Text);
dataOffset = outputDays;
double[] openPriceArray = new double[testOffset];
double[] closePriceArray = new double[testOffset];
double[] highPriceArray = new double[testOffset];;
double[] lowPriceArray = new double[testOffset];
double[] offsetArray = new double[testOffset];
for (int i = 1; i <= 10; i++)
{
ELM tempELM = new ELM(inputDays, hiddenLayerSize, outputDays, biasValue, data, trainPercentage);
switch (type)
{
case "Input Days":
tempELM = new ELM(i * testOffset, hiddenLayerSize, outputDays, biasValue, data, trainPercentage);
break;
case "Output Days":
tempELM = new ELM(inputDays, hiddenLayerSize, i * testOffset, biasValue, data, trainPercentage);
break;
case "Hidden Layer Size":
tempELM = new ELM(inputDays, i * testOffset, outputDays, biasValue, data, trainPercentage);
break;
case "Train Percentage":
tempELM = new ELM(inputDays, hiddenLayerSize, outputDays, biasValue, data, i * testOffset);
break;
}
tempELM.train(tempELM.Xtrain, tempELM.Ytrain);
Matrix <double> tempPrediction = tempELM.predict(tempELM.Xtest);
Matrix <double> tempTest = tempELM.Ytest;
int tempBestIndex = GetBestIndex(tempPrediction, tempTest);
double[] selectedPredSample = tempPrediction.Row(tempBestIndex).ToArray();
double[] selectedTestSample = tempTest.Row(tempBestIndex).ToArray();
double[] predOpenPrice = new ArraySegment <double>(selectedPredSample, 0, dataOffset).ToArray();
double[] predClosePrice = new ArraySegment <double>(selectedPredSample, dataOffset, dataOffset).ToArray();
double[] predHighPrice = new ArraySegment <double>(selectedPredSample, 2 * dataOffset, dataOffset).ToArray();
double[] predLowPrice = new ArraySegment <double>(selectedPredSample, 3 * dataOffset, dataOffset).ToArray();
double[] testOpenPrice = new ArraySegment <double>(selectedTestSample, 0, dataOffset).ToArray();
double[] testClosePrice = new ArraySegment <double>(selectedTestSample, dataOffset, dataOffset).ToArray();
double[] testHighPrice = new ArraySegment <double>(selectedTestSample, 2 * dataOffset, dataOffset).ToArray();
double[] testLowPrice = new ArraySegment <double>(selectedTestSample, 3 * dataOffset, dataOffset).ToArray();
double tempOpenPriceError = GetMeanSquare(predOpenPrice, testOpenPrice);
double tempClosePriceError = GetMeanSquare(predClosePrice, testClosePrice);
double tempHighPriceError = GetMeanSquare(predHighPrice, testHighPrice);
double tempLowPriceError = GetMeanSquare(predLowPrice, testLowPrice);
offsetArray[i - 1] = i * testOffset;
openPriceArray[i - 1] = tempOpenPriceError;
closePriceArray[i - 1] = tempClosePriceError;
highPriceArray[i - 1] = tempHighPriceError;
lowPriceArray[i - 1] = tempLowPriceError;
}
StatWindow inputStatWindow = new StatWindow(openPriceArray, closePriceArray, highPriceArray, lowPriceArray, offsetArray, type);
inputStatWindow.Show();
}