public Models.App.QtdResponse GetQtdResponse([FromBody] Models.App.QtdResponseBody qtdResponseBody)
{
Models.App.QtdResponse qtdResponse = new Models.App.QtdResponse();
Regressor regressor = new Regressor();
string filterTxt = JsonConvert.SerializeObject(qtdResponseBody.FilterOptions);
regressor.LoadByFilter(filterTxt);
string text = System.IO.File.ReadAllText(Location.ConfLocation);
Conf conf = JsonConvert.DeserializeObject <Conf>(text);
if (regressor.RegressorId == null || regressor.RegressorId == "")
{
//treinar
string Regressor_Id = Encoder.Base64Encode(Convert.ToString(DateTime.Now));
string trainerDir = conf.TrainerDir;
string Trainercmd = "python main.py ";
Trainercmd += qtdResponseBody.appId;
Trainercmd += " ";
Trainercmd += Regressor_Id;
Trainercmd += " ";
Trainercmd += filterTxt.Replace("\"", "\\\"");
Command.runCmd(trainerDir, Trainercmd);
Regressor insertRegressor = new Regressor();
insertRegressor.RegressorId = Regressor_Id;
insertRegressor.AppId = qtdResponseBody.appId;
insertRegressor.Filter = filterTxt;
insertRegressor.Insert();
}
//consultar
qtdResponse.Qtd = new List <double>();
string ConsultDir = conf.PredictorDir;
string ConsultCmd = "python main.py ";
ConsultCmd += qtdResponseBody.appId;
ConsultCmd += " ";
ConsultCmd += regressor.RegressorId;
ConsultCmd += " ";
ConsultCmd += qtdResponseBody.Min;
ConsultCmd += " ";
ConsultCmd += qtdResponseBody.Max;
ConsultCmd += " ";
ConsultCmd += qtdResponseBody.Step;
string Image = Command.runCmd(ConsultDir, ConsultCmd);
Models.App.Previsor previsor = JsonConvert.DeserializeObject <Models.App.Previsor>(Image);
qtdResponse.Qtd = previsor.Prev;
return(qtdResponse);
}
private void Start()
{
regressor = FindObjectOfType <Regressor>();
if (regressor == null)
{
Debug.LogError("Could not find Regressor!");
}
dataSet = regressor.dataSet;
if (dataSet == null)
{
Debug.LogError("Could not find dataSet on Regressor!");
}
}
public double Predict(double spot)
{
if (Regressor == null)
{
lock (_threadlock)
{
if (Regressor == null)
{
Regressor = PerformRegression();
}
}
}
return(Regressor.Interpolate(spot));
}
public static void SimpleExampleLoadFromString()
{
var trainFile = Path.Combine(_dir, "regression", "regression.train");
var testFile = Path.Combine(_dir, "regression", "regression.test");
var trainTpl = GetDataAndLabel(trainFile);
var testTpl = GetDataAndLabel(testFile);
var X_test = testTpl.Item1;
var y_test = testTpl.Item2;
var est = new Regressor()
{
BoostingType = "gbdt",
Metric = new[] { Metric.l2, Metric.auc },
NumLeaves = 31,
LearningRate = 0.05,
FeatureFraction = 0.9,
BaggingFraction = 0.8,
BaggingFreq = 5,
Verbose = 1,
EarlyStoppingRound = 5,
NumIterations = 20
};
var estDict = est.ToDictionary();
var dsTrain = new Dataset(trainFile);
var dsEval = new Dataset(testFile, reference: dsTrain);
var res = new Dictionary <string, Dictionary <string, List <double> > >();
string modelString;
using (var bst = LightGBMSharp.LGBM.Train(est, dsTrain, res, dsEval))
{
Console.WriteLine("Save model...");
bst.SaveModel("model.txt");
modelString = bst.SaveModelToString();
}
using (var bst2 = Booster.LoadFromModelString(modelString))
{
Console.WriteLine("Start predicting...");
var y_pred = bst2.Predict(X_test, LGBMPredictType.PredictNormal);
var mse = Distance.MSE(y_test, y_pred);
var y_predS = String.Join(", ", y_pred.Take(10));
Console.WriteLine($"Pred == {y_predS}");
Console.WriteLine($"The rmse of prediction is: {Math.Sqrt(mse)}");
}
}
public static void SimpleExample()
{
var trainFile = Path.Combine(_dir, "regression", "regression.train");
var testFile = Path.Combine(_dir, "regression", "regression.test");
//var testMatrix = DelimitedReader.Read<double>(testFile, false, "\t", false, CultureInfo.InvariantCulture);
//var y_test = testMatrix.Column(0).ToArray();
//var X_test = testMatrix.RemoveColumn(0).ToArray();
var trainTpl = GetDataAndLabel(trainFile);
var testTpl = GetDataAndLabel(testFile);
var X_test = testTpl.Item1;
var y_test = testTpl.Item2;
var est = new Regressor()
{
BoostingType = "gbdt",
Metric = new[] { Metric.l2, Metric.auc },
NumLeaves = 31,
LearningRate = 0.05,
FeatureFraction = 0.9,
BaggingFraction = 0.8,
BaggingFreq = 5,
Verbose = 1,
EarlyStoppingRound = 5,
NumIterations = 20
};
var estDict = est.ToDictionary();
var dsTrain = new Dataset(trainFile);
var dsEval = new Dataset(testFile, reference: dsTrain);
var res =
new Dictionary <string, Dictionary <string, List <double> > >();
var bst = LightGBMSharp.LGBM.Train(est, dsTrain, res, dsEval);
Console.WriteLine("Save model...");
bst.SaveModel("model.txt");
Console.WriteLine("Start predicting...");
var y_pred = bst.Predict(X_test, LGBMPredictType.PredictNormal);
var mse = Distance.MSE(y_test, y_pred);
var y_predS = String.Join(", ", y_pred.Take(10));
Console.WriteLine($"Pred == {y_predS}");
Console.WriteLine($"The rmse of prediction is: {Math.Sqrt(mse)}");
}
public string TrainRegressor([FromQuery(Name = "Id")] string Id, [FromQuery(Name = "Hash")] string Hash)
{
string text = System.IO.File.ReadAllText(Location.ConfLocation);
Conf conf = JsonConvert.DeserializeObject <Conf>(text);
Regressor regressor = new Regressor();
regressor.LoadByHash(Hash);
string trainerDir = conf.TrainerDir;
string Trainercmd = "python main.py ";
Trainercmd += Id;
Trainercmd += " ";
Trainercmd += regressor.RegressorId;
Trainercmd += " ";
Trainercmd += regressor.Filter.Replace("\"", "\\\"");
Command.runCmd(trainerDir, Trainercmd);
return("ok");
}
/// <summary>
/// Perform target alignment
/// </summary>
/// <param name="evidences">Evidences to align</param>
/// <param name="baseline">Baseline evidences</param>
/// <returns></returns>
public LinearRegressionResult AlignTargets(List <Evidence> evidences, List <Evidence> baseline)
{
var observed = new List <double>();
var predicted = new List <double>();
foreach (var evidence in baseline)
{
Predictor = RetentionTimePredictorFactory.CreatePredictor(Options.PredictorType);
evidence.PredictedNet = Predictor.GetElutionTime(Evidence.CleanSequence(evidence.Sequence));
observed.Add(evidence.ObservedNet);
predicted.Add(evidence.PredictedNet);
}
Regressor = LinearRegressorFactory.Create(Options.RegressionType);
var result = Regressor.CalculateRegression(observed, predicted);
foreach (var evidence in evidences)
{
evidence.ObservedNet = Regressor.Transform(result, evidence.ObservedNet);
}
return(result);
}
