//Todo: Create dialog
private void optimizeParametersNN_btn_Click(object sender, EventArgs e)
{
new Thread(delegate() {
//Todo: Drop fields random
string[] inputFields = new string[] { "ssi-mt4", "spread", "mid-TradingTime", "mid-Stoch_600000", "mid-Stoch_1800000", "mid-Stoch_3600000", "mid-Stoch_7200000", "mid-Stoch_14400000", "mid-Stoch_21600000", "mid-MA_600000", "mid-MA_1800000", "mid-MA_3600000", "mid-MA_7200000", "mid-MA_14400000", "mid-MA_21600000", "mid-Range_1800000", "mid-Range_3600000", "mid-Range_7200000" };
string outputField = "buy-outcomeCode-0,001_600000";
int epochs = 10;
int networksCount = 500;
string path = Application.StartupPath + "/AI/";
GeneralExcelGenerator excel = new GeneralExcelGenerator(path + "excel.xlsx");
string sheetName = "OptimizeNN-E10";
string[] header = new string[2 + AdvancedNeuralNetwork.getExcelHeader().Length];
AdvancedNeuralNetwork.getExcelHeader().CopyTo(header, 2);
header[0] = "ValidationE";
header[1] = "TrainingE";
excel.CreateSheet(sheetName, header);
double[][] inputsTraining = new double[][] { };
double[][] outputsTraining = new double[][] { };
dataminingDb.getInputOutputArrays(inputFields, outputField, "EURUSD", ref inputsTraining, ref outputsTraining, DataGroup.All, 100 * 1000, 0);
double[][] inputsValidation = new double[][] { };
double[][] outputsValidation = new double[][] { };
dataminingDb.getInputOutputArrays(inputFields, outputField, "EURUSD", ref inputsValidation, ref outputsValidation, DataGroup.All, 10 * 1000, 1);
dataminingDb.unloadPair("EURUSD");
for (int ns = 0; ns < networksCount; ns++)
{
IMachineLearning network = AdvancedNeuralNetwork.getRandom(inputFields, outputField);
network.train(inputsTraining, outputsTraining, epochs);
double trainingError = network.getError();
double validationError = network.getPredictionErrorFromData(inputsValidation, outputsValidation);
string[] excelRow = new string[2 + network.getInfo(inputFields, outputField).Length];
network.getInfo(inputFields, outputField).CopyTo(excelRow, 2);
excelRow[0] = validationError.ToString();
excelRow[1] = trainingError.ToString();
excel.addRow(sheetName, excelRow);
setState("Optimizing NN " + ns + "/" + networksCount);
}
if (Directory.Exists(path) == false)
{
Directory.CreateDirectory(path);
}
excel.FinishSheet(sheetName);
excel.FinishDoc();
excel.save();
}).Start();
}
public double TestMachineLearning(IMachineLearning learning)
{
double[][] input = new double[10][];
input[0] = new double[] { 1, 2, 1 };
input[1] = new double[] { 1, 2, 1 };
input[2] = new double[] { 1, 2, 2 };
input[3] = new double[] { 2, 2, 1 };
input[4] = new double[] { 2, 2, 3 };
input[5] = new double[] { 3, 2, 2 };
input[6] = new double[] { 3, 2, 1 };
input[7] = new double[] { 3, 2, 5 };
input[8] = new double[] { 4, 3, 3 };
input[9] = new double[] { 4, 3, 2 };
double[][] output = new double[10][];
output[0] = new double[] { 1, 0 };
output[1] = new double[] { 1, 0 };
output[2] = new double[] { 1, 0 };
output[3] = new double[] { 1, 1 };
output[4] = new double[] { 1, 1 };
output[5] = new double[] { 1, 1 };
output[6] = new double[] { 0, 1 };
output[7] = new double[] { 0, 1 };
output[8] = new double[] { 0, 1 };
output[9] = new double[] { 0, 1 };
learning.train(input, output, 1);
double[] low = learning.getPrediction(new double[] { 0, 0, 0 });
double[] mid = learning.getPrediction(new double[] { 2, 2, 2 });
double[] high = learning.getPrediction(new double[] { 5, 5, 5 });
double score = Math.Abs(low[0] - 1) + Math.Abs(low[1] - 0) +
Math.Abs(mid[0] - 1) + Math.Abs(mid[1] - 1) +
Math.Abs(high[0] - 0) + Math.Abs(high[1] - 1);
score /= 6;
Assert.IsTrue(false, "Low: " + low[0] + " " + low[1] + Environment.NewLine + "Mid: " + mid[0] + " " + mid[1] + Environment.NewLine + "Heigh: " + high[0] + " " + high[1]);
return(score);
}
