private void showTree(IChromosome ch, Parameters param)
{
if (ch == null)
{
return;
}
var model = (GPdotNet.Core.Chromosome)ch;
if (model == null)
{
return;
}
m_bestSolution = ch;
Parameters = param;
//
solutionExpression.Text = NodeDecoding.Decode(model.expressionTree, param, BasicTypes.EncodeType.Default);
if (treeCtrlDrawer1 == null)
{
initTreeDrawer();
}
treeCtrlDrawer1.DrawTreeExpression(model.expressionTree, Globals.FunctionFromId, param);
//show constance
showConstances(param);
}
public static void Export(Node ch, Parameters param, IData data, Experiment experiment, string strFilePath)
{
try
{
// open selected file and retrieve the content
using (TextWriter tw = new StreamWriter(strFilePath))
{
tw.Flush();
//Add Data.
var cols = experiment.GetColumnsFromInput();
var outCol = experiment.GetColumnsFromOutput(false).FirstOrDefault();
string cmd = "training_data_set={";
for (int i = 0; i < experiment.GetRowCount(false); i++)
{
string line = "{";
//get normalized and numeric row
var row = experiment.GetRowFromInput(i, false);
var row_norm = experiment.GetEncodedInput(i, false, false);
//input columns
for (int j = 0; j < row_norm.Length; j++)
{
line += row_norm[j].ToString(CultureInfo.InvariantCulture);
if (j + 1 != row_norm.Length)
{
line += ",";
}
else
{
if (outCol.ColumnDataType == ColumnType.Category)
{
line += "," + outCol.GetNumericValue(i).Value.ToString(CultureInfo.InvariantCulture);
}
else
{
line += "," + experiment.GetRowFromOutput(i, false)[0].ToString(CultureInfo.InvariantCulture);
}
line += "}";
}
//
}
cmd += line;
if (i + 1 < experiment.GetRowCount(false))
{
cmd += ",";
}
else
{
cmd += "};";
}
}
tw.WriteLine(cmd);
//GP Model formula
string formula = NodeDecoding.Decode(ch, param, EncodeType.Mathematica);
List <string> inputArgs = new List <string>();
AlphaCharEnum alphaEnum = new AlphaCharEnum();
var totCols = experiment.GetEncodedColumnInputCount();
var diff = totCols - cols.Count;//difference between column count and normalized column count due to Category column clusterization
for (int i = totCols - 1; i >= 0; i--)
{
string var = "x" + (i + 1).ToString() + " ";
//make a formula to de normalize value
var col = cols[i - diff];
if (col.ColumnDataType == ColumnType.Category)
{
//formula = formula.Replace(var, replCell);
if (diff > 0)
{
diff -= 1;
}
}
else if (col.ColumnDataType == ColumnType.Binary)
{
//formula = formula.Replace(var, replCell);
}
else
{
var replCell = GPToExcel.createNormalizationFormulaForColumn(col, var);
formula = formula.Replace(var, replCell);
}
//
inputArgs.Add(var);
}
//Replace random constants with real values
var consts = data.GetRandConsts();
for (int i = 0; i < consts.Length; i++)
{
string var = "r" + (i + 1).ToString() + " ";
string constValue = consts[i].ToString(CultureInfo.InvariantCulture);
if (constValue[0] == '-')
{
constValue = "(" + constValue + ")";
}
formula = formula.Replace(var, constValue);
}
//in case of category output
//category output is pre calculated with sigmoid multiply with Class count.
if (outCol.ColumnDataType == ColumnType.Numeric)//for numeric output we need to de normalize formula
{
var normFormula = GPToExcel.createDeNormalizationFormulaForOutput(outCol, formula);
formula = normFormula;
}
//in case of softMax we must defined separate function in R
var customFun = "";
if (param.RootFunctionNode != null && param.RootFunctionNode.Name == "Softmax")
{
customFun =
"Softmax[x_List] := Ordering[Exp[x - Max[x]] / Total[Exp[x - Max[x]]], 1][[1]] -1; " + Environment.NewLine;
}
//add model name and arguments
formula = "gpModel[{0}]:=" + formula;
//add arguments to the model
string arguments = "";
for (int i = 0; i < inputArgs.Count; i++)
{
var a = inputArgs[i];
if (formula.Contains(a))
{
if (i == 0)
{
a = a.Replace(" ", "_");
}
else
{
a = a.Replace(" ", "_,");
};
//
arguments = a + arguments;
}
}
if (arguments.EndsWith(","))
{
arguments = arguments.Substring(0, arguments.Length - 1);
}
formula = string.Format(formula, arguments);
tw.WriteLine(formula + ";");
tw.Close();
}
}
catch (Exception)
{
throw;
}
}
public static void Export(Node ch, Parameters param, IData data, Experiment experiment, string strFilePath)
{
try
{
// open selected file and retrieve the content
using (TextWriter tw = new StreamWriter(strFilePath))
{
tw.Flush();
//GP Model formula
//
string formula = NodeDecoding.Decode(ch, param, EncodeType.RLanguage);
List <string> inputArgs = new List <string>();
AlphaCharEnum alphaEnum = new AlphaCharEnum();
//make a formula to de normalize value
var cols = experiment.GetColumnsFromInput();
var totCols = experiment.GetEncodedColumnInputCount();
var diff = totCols - cols.Count;//difference between column count and normalized column count due to Category column clusterization
///
for (int i = totCols - 1; i >= 0; i--)
{
string var = "x" + (i + 1).ToString() + " ";
string cell = var;
//make a formula to de normalize value
var col = cols[i - diff];
string replCell = cell;
if (col.ColumnDataType == ColumnType.Category)
{
formula = formula.Replace(var, replCell);
if (diff > 0)
{
diff -= 1;
}
}
else if (col.ColumnDataType == ColumnType.Binary)
{
formula = formula.Replace(var, replCell);
}
else
{
replCell = GPToExcel.createNormalizationFormulaForColumn(col, cell);
formula = formula.Replace(var, replCell);
}
//
inputArgs.Add(var);
}
//Replace random constants with real values
var consts = data.GetRandConsts();
for (int i = 0; i < consts.Length; i++)
{
string var = "r" + (i + 1).ToString() + " ";
string constValue = consts[i].ToString(CultureInfo.InvariantCulture);
formula = formula.Replace(var, constValue);
}
//in case of category output
//category output is pre-calculated with sigmoid multiply with Class count.
var outCol = experiment.GetColumnsFromOutput().FirstOrDefault();
//de normalize output column if available
if (outCol.ColumnDataType == ColumnType.Numeric)
{
var normFormula = GPToExcel.createDeNormalizationFormulaForOutput(outCol, formula);
formula = normFormula;
}
//in case of softMax we must defined separate function in R
var customFun = "";
if (param.RootFunctionNode != null && param.RootFunctionNode.Name == "Softmax")
{
customFun =
"Softmax <- function(x) {" + Environment.NewLine +
" max = max(x)" + Environment.NewLine +
" sum = sum(exp(x - max))" + Environment.NewLine +
" which.max(exp(x - max) / sum)-1;" + Environment.NewLine +
" };" + Environment.NewLine;
}
//create training and testing data frame
var trainingDataset = createDataFrame(experiment, false);
var testingDataset = createDataFrame(experiment, true);
//add arguments to the model
string arguments = "";
string calcArguments = "";
for (int i = 0; i < inputArgs.Count; i++)
{
int k = inputArgs.Count - i - 1;
var a = inputArgs[k];
if (formula.Contains(a))
{
a = a.Replace(" ", ",");
//
arguments = arguments + a;
calcArguments += $"x[{i + 1}],";
}
}
if (!string.IsNullOrEmpty(arguments) && arguments.EndsWith(","))
{
arguments = arguments.Substring(0, arguments.Length - 1);
}
//calculate output column for training and testing data set
var calcTrain = "training_set$Ygp <- apply(training_set, 1, function(x) gpModel(" + calcArguments.Substring(0, calcArguments.Length - 1) + "));"; // x[1], x[2], x[3], x[4]));";
var calcTest = "testing_set$Ygp <- apply(testing_set, 1, function(x) gpModel(" + calcArguments.Substring(0, calcArguments.Length - 1) + "));"; // x[1], x[2], x[3], x[4]));"
//construct function
formula = @"gpModel<- function(" + arguments + ") {" + Environment.NewLine + formula + Environment.NewLine + "};";
//add custom defined function
formula = trainingDataset + Environment.NewLine + testingDataset + Environment.NewLine + customFun + formula +
Environment.NewLine + calcTrain + Environment.NewLine + calcTest;
tw.WriteLine(formula);
tw.Close();
}
}
catch (Exception)
{
throw;
}
}
/// <summary>
/// Exports data and the model to excel
/// </summary>
/// <param name="experiment"></param>
/// <param name="inputVarCount"></param>
/// <param name="constCount"></param>
/// <param name="ch"></param>
/// <param name="strFilePath"></param>
public static void Export(Node ch, Parameters param, IData data, Experiment experiment, string strFilePath)
{
try
{
//
var wb = new XLWorkbook();
var ws1 = wb.Worksheets.Add("TRAINING DATA");
var ws2 = ws1;
if (experiment.IsTestDataExist())
{
ws2 = wb.Worksheets.Add("TESTING DATA");
}
else
{
ws2 = null;
}
ws1.Cell(1, 1).Value = "Training Data";
if (experiment.IsTestDataExist())
{
ws2.Cell(1, 1).Value = "Testing Data";
}
writeDataToExcel(experiment, ws1, false);
if (experiment.IsTestDataExist())
{
writeDataToExcel(experiment, ws2, true);
}
//GP Model formula
string formula = NodeDecoding.Decode(ch, param, EncodeType.Excel);
AlphaCharEnum alphaEnum = new AlphaCharEnum();
//make a formula to de normalize value
var cols = experiment.GetColumnsFromInput();
var totCols = experiment.GetEncodedColumnInputCount();
var diff = totCols - cols.Count;//difference between column count and normalized column count due to Category column clusterization
//Replace Input variable with excel column references
for (int i = totCols - 1; i >= 0; i--)
{
string var = "x" + (i + 1).ToString() + " ";
string cell = alphaEnum.AlphabetFromIndex(2 + i) + "3";
//make a formula to de normalize value
var col = cols[i - diff];
string replCell = cell;
if (col.ColumnDataType == ColumnType.Category)
{
formula = formula.Replace(var, replCell);
if (diff > 0)
{
diff -= 1;
}
}
else if (col.ColumnDataType == ColumnType.Binary)
{
formula = formula.Replace(var, replCell);
}
else
{
replCell = createNormalizationFormulaForColumn(col, cell);
formula = formula.Replace(var, replCell);
}
}
//Replace random constants with real values
int constCount = 1;
var consts = data.GetRandConsts();
for (int i = 0; i < consts.Length; i++)
{
string var = "r" + (i + 1).ToString() + " ";
string constValue = consts[i].ToString();
formula = formula.Replace(var, constValue);
constCount++;
}
//output column
var outCol = experiment.GetColumnsFromOutput().FirstOrDefault();
//de normalize output column if available
if (outCol.ColumnDataType == ColumnType.Numeric)
{
var normFormula = createDeNormalizationFormulaForOutput(outCol, formula);
formula = normFormula;
}
//in case of decimal point, semicolon of Excel formula must be replaced with comma
if ("." == CultureInfo.CurrentCulture.NumberFormat.NumberDecimalSeparator)
{
formula = formula.Replace(";", ",");
}
ws1.Cell(3, totCols + 3).Value = formula;
if (experiment.IsTestDataExist())
{
ws2.Cell(3, totCols + 3).Value = formula;
}
//
wb.SaveAs(strFilePath, false);
}
catch (Exception)
{
throw;
}
}
