//**************************************************************************************
public static WekaLinearRegression CreateNew(int iParamtersID, List <ClassifierParameter> iParameters, Candlestick.Period iPeriod, int iProfitTime, List <int> iTrainingPoints, bool iEliminateColinearAttributes, bool iMinimal)
{
// Create WekaInfo and retrive ID
var info = new WekaInfo(null, typeof(WekaJ48), iParamtersID, iPeriod, iProfitTime);
WekaInfo.InsertToDB(info);
if (info.ID is null)
{
throw new Exception("Could not deduct ID");
}
// Create LinearRegressionInfo
var lrInfo = new WekaLinearRegressionInfo((int)info.ID, iEliminateColinearAttributes, iMinimal);
WekaLinearRegressionInfo.UpdateDB(lrInfo);
// Save training points
if (iTrainingPoints != null)
{
SaveTrainingPoints(iTrainingPoints, "Data/TrainingPoints/Classifier_" + info.ID.ToString() + ".dat");
}
// Create classifier
return(new WekaLinearRegression(info, iParameters, lrInfo));
}
//**************************************************************************************
public override void RebuildClassifier()
{
OutputMessage("Rebuilding classifier");
// Remove model
DeleteModel(ID);
// Remove training data
DeleteTrainingPoints(ID);
// Update info
LRInfo.OnClassifierRebuild();
WekaLinearRegressionInfo.UpdateDB(LRInfo);
}
//**************************************************************************************
/// <summary>
/// Build cllasifier model and save it to a file.
/// </summary>
public override void Build(CandlestickCollection iCandlestick)
{
List <int> trainingPoints = null;
// Build model
if (!File.Exists(ModelFilename))
{
OutputMessage("Building model");
trainingPoints = LoadTrainingPoints(iCandlestick, ID, ProfitTime);
Model = new weka.classifiers.functions.LinearRegression();
Model.setEliminateColinearAttributes(LRInfo.EliminateColinearAttributes);
Model.setMinimal(LRInfo.Minimal);
Model.buildClassifier(CreateInstances(iCandlestick, trainingPoints, Attributes, Parameters, Period, ProfitTime));
weka.core.SerializationHelper.write(ModelFilename, Model);
coefficients = Model.coefficients();
}
// Calculate coeficient of determination and profits std dev
if (LRInfo.CoefOfDetermination == null || LRInfo.ProfitStdDev == null)
{
WaitTillModelReady();
OutputMessage("Calculating coeficient of determination and profits std dev");
if (trainingPoints is null)
{
trainingPoints = LoadTrainingPoints(iCandlestick, ID, ProfitTime);
}
var par = CalculateParameters(Parameters, iCandlestick, trainingPoints, Period);
var futureProfits = FullToTraining(new List <float>(CalculateFutureProfits(iCandlestick[kTrainingPeriod], ProfitTime)), trainingPoints).ToArray();
var pred = new List <float>();
for (int j = 0; j < par.Count; j++)
{
var fp = PredictDFP(par[j]);
var targetPriceChange = (float)Math.Exp(fp);
pred.Add(targetPriceChange - 1.0f);
}
LRInfo.CoefOfDetermination = (float)Math.Pow(Statistics.CalculateCorrelation(futureProfits, pred.ToArray()), 2.0f);
LRInfo.ProfitStdDev = Statistics.StandardDeviation(futureProfits, pred.ToArray());
WekaLinearRegressionInfo.UpdateDB(LRInfo);
}
}