//public double[][] MainProcess()
//{
// for (int i = 0; i < NumPattern; i++)
// {
// ANNModel.OutInputLayer = ArrPattern[i];
// DblActual_Forecast[1][i] = ANNModel.FeedforwardTraining();
// }
// DblActual_Forecast[0] = Preprocess.DenormalizeByMinMax(DblActual_Forecast[0]);
// DblActual_Forecast[1] = Preprocess.DenormalizeByMinMax(DblActual_Forecast[1]);
// return DblActual_Forecast;
//}
public double[][] MainProcessTrend()
{
double[] dblTempForecast;
for (int i = 0; i < NumPattern; i++)
{
ANNModel.OutInputLayer = ArrPattern[i];
dblTempForecast = ANNModel.FeedforwardTraining();
//Xử lý kết quả dự đoán từ 3 node về xu hướng xác định
if (dblTempForecast[0] > dblTempForecast[1] && dblTempForecast[0] > dblTempForecast[2])
{
DblActual_Forecast[1][i] = -1;// DOWNTREND
}
else if ((dblTempForecast[1] > dblTempForecast[2] && dblTempForecast[1] > dblTempForecast[0]))
{
DblActual_Forecast[1][i] = 0;// NOTREND
}
else
{
DblActual_Forecast[1][i] = 1;// UPTREND
}
}
//DblActual_Forecast[0] = Preprocess.DenormalizeByMinMax(DblActual_Forecast[0]);
//DblActual_Forecast[1] = Preprocess.DenormalizeByMinMax(DblActual_Forecast[1]);
return(DblActual_Forecast);
}
/// <summary>
/// Deserilization of run condition
/// </summary>
/// <param name="p"></param>
public void ActivatePanel(ANNModel model)
{
//prepare graphs
PrepareGraphs(model);
//update model
updateChartData(model.ActualV, model.PredictedV);
this.zedModel.RestoreScale(zedModel.GraphPane);
}
/// <summary>
/// Deserilization of run condition
/// </summary>
/// <param name="p"></param>
public void ActivatePanel(ActiveModelData data, ANNModel model)
{
SetParameters(data);
//clear previous data if exist
lossFunction.Clear();
evalFunciton.Clear();
actualData.Clear();
predictedData.Clear();
//prepare graphs
PrepareGraphs(model);
//update loss and eval
for (int i = 0; i < model.LossData.Count; i++)
{
updateChartFitness(i, model.LossData[i], model.EvalData[i], false);
}
//update mdoel
updateChartData(model.ActualT, model.PredictedT);
zedFitness.RestoreScale(zedFitness.GraphPane);
this.zedModel.RestoreScale(zedModel.GraphPane);
}
public void Show()
{
var page1 = new WizardPage();
page1.hasBackButton = false;
page1.hasCancelButton = true;
page1.hasNextButton = true;
page1.hasFinishButton = false;
page1.cancelAction = () =>
{
page1.Close();
};
page1.Title = "Neural Network Model Wizard";
page1.HeaderTitle = "Step 1 - Model Management";
page1.HeaderDescription = "Select an Action";
page1.FooterText = "Click 'Next' to Continue.";
var dl = c.GetDefaultContainer();
dl.Width = Width;
dl.CreateAndAddLabelRow2("Select an action:");
dl.CreateAndAddEmptySpace();
dl.CreateAndAddEmptySpace();
var rl = new RadioButtonList {
Orientation = Orientation.Vertical
};
rl.Spacing = new Size(10, 10);
rl.Items.Add("Create and Train a New Model");
rl.Items.Add("Load an Existing Model");
rl.SelectedIndex = 0;
page1.nextAction = () =>
{
switch (rl.SelectedIndex)
{
case 0:
CurrentModel = new ANNModel();
Application.Instance.Invoke(() => DisplayPage_LoadData());
break;
case 1:
CurrentModel = new ANNModel();
Application.Instance.Invoke(() => DisplayPage_LoadModel());
break;
}
page1.Close();
};
dl.CreateAndAddControlRow(rl);
var scrollable = new Scrollable {
Content = dl, Border = BorderType.None, Size = new Size(Width, Height), ExpandContentHeight = true
};
page1.Init(Width, Height);
page1.Topmost = false;
page1.ContentContainer.Add(scrollable);
page1.Show();
}
void Init()
{
nf = SimObject.FlowSheet.FlowsheetOptions.NumberFormat;
su = SimObject.FlowSheet.FlowsheetOptions.SelectedUnitSystem;
CurrentModel = new ANNModel();
}
/// <summary>
/// Initi props for charts
/// </summary>
public void PrepareGraphs(ANNModel model)
{
//clear prev. state
zedModel.GraphPane.CurveList.Clear();
//
//
predictedData = new LineItem("Predicted", null, null, Color.Blue, ZedGraph.SymbolType.None, 1);
predictedData.Symbol.Fill = new Fill(Color.Blue);
predictedData.Symbol.Size = 1;
//
actualData = new LineItem("Actual", null, null, Color.Red, ZedGraph.SymbolType.None, 1);
actualData.Symbol.Fill = new Fill(Color.Red);
actualData.Symbol.Size = 1;
//
if (model.OutputDim == 1)
{
predictedData = new LineItem("Predicted", null, null, Color.Blue, ZedGraph.SymbolType.None, 1);
predictedData.Symbol.Fill = new Fill(Color.Blue);
predictedData.Symbol.Size = 1;
//
actualData = new LineItem("Actual", null, null, Color.Red, ZedGraph.SymbolType.None, 1);
actualData.Symbol.Fill = new Fill(Color.Red);
actualData.Symbol.Size = 1;
zedModel.GraphPane.YAxis.Scale.MaxAuto = true;
zedModel.GraphPane.YAxis.Scale.MinAuto = true;
zedModel.GraphPane.YAxis.MajorGrid.IsVisible = false;
zedModel.GraphPane.XAxis.MajorGrid.IsVisible = false;
zedModel.GraphPane.YAxis.Scale.MajorStepAuto = true;
zedModel.GraphPane.YAxis.Scale.MinorStepAuto = true;
zedModel.GraphPane.XAxis.Scale.MajorStep = 1;
zedModel.GraphPane.XAxis.Scale.MinorStep = 0;
}
else
{
Classes = model.Classes;
max = model.Classes.Count - 1;
predictedData = new LineItem("Predicted", null, null, Color.Red, ZedGraph.SymbolType.Circle, 0);
predictedData.Symbol.Fill = new Fill(Color.Red);
predictedData.Symbol.Size = 5;
actualData = new LineItem("Actual", null, null, Color.Blue, ZedGraph.SymbolType.Diamond, 0);
actualData.Symbol.Border = new Border(Color.Blue, 0.5f);
actualData.Symbol.Fill = new Fill(Color.Blue);
actualData.Symbol.Size = 4;
zedModel.GraphPane.YAxis.ScaleFormatEvent += YAxis_ScaleFormatEvent;
//zedModel.GraphPane.XAxis.Scale.Min = 0;
zedModel.GraphPane.YAxis.Scale.Min = -0.1;
zedModel.GraphPane.YAxis.Scale.Max = max + 0.2;
zedModel.GraphPane.YAxis.MajorGrid.DashOff = 0;
zedModel.GraphPane.YAxis.MajorGrid.IsVisible = true;
zedModel.GraphPane.XAxis.MajorGrid.DashOff = 0;
zedModel.GraphPane.XAxis.MajorGrid.IsVisible = false;
zedModel.GraphPane.YAxis.Scale.MajorStep = 1;
zedModel.GraphPane.YAxis.Scale.MinorStep = 1;
zedModel.GraphPane.XAxis.Scale.MajorStep = 1;
zedModel.GraphPane.XAxis.Scale.MinorStep = 1;
}
zedModel.GraphPane.Title.Text = "ANN Iteration Simulation";
zedModel.GraphPane.XAxis.Title.Text = "Samples";
zedModel.GraphPane.YAxis.Title.Text = model.Label;
zedModel.GraphPane.YAxis.MajorGrid.IsZeroLine = false;
zedModel.GraphPane.CurveList.Add(actualData);
zedModel.GraphPane.CurveList.Add(predictedData);
}
public void Reset(ANNModel model)
{
ResetChart();
PrepareGraphs(model);
}
public void Reset(ANNModel model)
{
ResetChart();
PrepareGraphs(model);
SetParameters(ActiveModelData.GetDefaults());
}
public void Run(ANNModel model, ActiveModelData setData, Action <int, float, float, (List <List <float> >, List <List <float> >, List <List <float> >), (List <List <float> >, List <List <float> >, List <List <float> >)> report, CancellationToken token)
//public void LoadDataSet(double[][] trainingSet)
//{
// NumPattern = trainingSet.Length;
// ArrActual = new double[NumPattern][];
// ArrPredict = new double[NumPattern][];
// ArrPattern = new double[NumPattern][];
// for (int i = 0; i < NumPattern; i++)
// {
// ArrActual[i] = trainingSet[i][ANNParameterBUS.InputNode];
// ArrPattern[i] = new double[ANNParameterBUS.InputNode];
// for (int j = 0; j < ANNParameterBUS.InputNode; j++)
// {
// ArrPattern[i][j] = trainingSet[i][j];
// }
// }
// ANNModel = new ANNModelBUS();
//}
//public void Main()
//{
// //Khởi tạo bộ trọng số ngẫu nhiên cho mô hình
// ANNModel.InitialWeight();
// TSCFDCostFunction();
// int loop = 1;
// bool bContinue = true;
// MeasureBUS measureBUS = new MeasureBUS();
// while (bContinue)
// {
// for (int i = 0; i < NumPattern; i++)
// {
// ANNModel.OutInputLayer = ArrPattern[i];
// ArrPredict[i] = ANNModel.FeedforwardTraining();
// }
// Error = measureBUS.NMSE(ArrActual, ArrPredict);
// if (Error <= 0.01 || loop >= ANNParameterBUS.MaxEpoch)
// {
// bContinue = false;
// ANNModel.SaveModelFile();
// //ghi nhận độ lỗi và số vòng lặp của quá trình train
// ANNModel.SaveError_MaxLoop(Error, loop);
// }
// else
// {
// for (int i = 0; i < NumPattern; i++)
// {
// ANNModel.OutInputLayer = ArrPattern[i];
// ANNModel.FeedforwardTraining();
// ANNModel.ErrorBackpropagationTraining(ArrDesiredValue[i]);
// }
// }
// loop++;
// }
//}
public void Main(int measureType)
{
//Khởi tạo bộ trọng số ngẫu nhiên cho mô hình
ANNModel.InitialWeight();
//TSCFDCostFunction();
int loop = 1;
bool bContinue = true;
MeasureBUS measureBUS = new MeasureBUS();
while (bContinue)
{
for (int i = 0; i < NumPattern; i++)
{
ANNModel.OutInputLayer = ArrPattern[i];
ArrPredicts[i] = ANNModel.FeedforwardTraining();
}
switch (measureType)
{
case MSE:
Convert2CaculateErrorMeasure();
Error = measureBUS.MSE(ArrTempActuals, ArrTempPredicts);
break;
case NMSE:
Convert2CaculateErrorMeasure();
Error = measureBUS.NMSE(ArrTempActuals, ArrTempPredicts);
break;
case RMSE:
Convert2CaculateErrorMeasure();
Error = measureBUS.RMSE(ArrTempActuals, ArrTempPredicts);
break;
case SIGN:
Convert2CaculateErrorMeasure();
Error = measureBUS.Sign(ArrTempActuals, ArrTempPredicts);
break;
}
bool bResult = CompareError_Accurancy(Error, ANNParameterBUS.Accuracy, measureType);
if (bResult == true || loop >= ANNParameterBUS.MaxEpoch)
{
bContinue = false;
ANNModel.SaveModelFile();
//ghi nhận độ lỗi và số vòng lặp của quá trình train
ANNModel.SaveError_MaxLoop(Error, loop);
}
else
{
for (int i = 0; i < NumPattern; i++)
{
ANNModel.OutInputLayer = ArrPattern[i];
ANNModel.FeedforwardTraining();
ANNModel.ErrorBackpropagationTraining(ArrActuals[i]);
}
}
loop++;
}
}
