static OHCL[] loadData(DateTime max)
{
if (max == null)
{
DateTime.Now.AddDays(1);
}
String jsonAsText = System.IO.File.ReadAllText(@"Z:\temp\BTCUSDT.txt");
Newtonsoft.Json.Linq.JContainer json = (Newtonsoft.Json.Linq.JContainer)JsonConvert.DeserializeObject(jsonAsText);
List <OHCL> array = new List <OHCL>();
foreach (var item in json)
{
try
{
OHCL ohcl = new OHCL();
ohcl.CurrentDate = UnixTimeStampToDateTime(float.Parse(item[0].ToString().Replace(".", ",").Substring(0, 10)));
if (ohcl.CurrentDate > max)
{
break;
}
ohcl.PreviousOpen = float.Parse(item.Previous[1].ToString().Replace(".", ","));
ohcl.PreviousHigh = float.Parse(item.Previous[2].ToString().Replace(".", ","));
ohcl.PreviousLow = float.Parse(item.Previous[3].ToString().Replace(".", ","));
ohcl.NextClose = float.Parse(item[4].ToString().Replace(".", ","));
ohcl.PreviousClose = float.Parse(item.Previous[4].ToString().Replace(".", ","));
ohcl.PreviousVolume = float.Parse(item.Previous[5].ToString().Replace(".", ","));
ohcl.PreviousQuoeteAssetVolume = float.Parse(item.Previous[7].ToString().Replace(".", ","));
ohcl.PreviousTrades = int.Parse(item.Previous[8].ToString().Replace(".", ","));
array.Add(ohcl);
}
catch
{
}
}
return(array.ToArray());
}
public static void start()
{
String lines = "";
try
{
OHCL[] list = loadData(DateTime.Now);
String l = "";
foreach (var item in list)
{
l += item.CurrentDate.ToString("yyyyMMdd") + ";" + item.NextClose + Environment.NewLine;
}
DateTime dateBegin = DateTime.Parse("2019-10-01");
String longOrShort = "null";
while (dateBegin < DateTime.Now.AddDays(-1))
{
//Import context
MLContext mlContext = new MLContext();
// 1. Import or create training data
OHCL[] ohclData = loadData(dateBegin);
IDataView trainingData = mlContext.Data.LoadFromEnumerable(ohclData);
// 2. Specify data preparation and model training pipeline
var pipeline = mlContext.Transforms.CopyColumns(outputColumnName: "Label", inputColumnName: "NextClose")
.Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "CurrentDateE", inputColumnName: "CurrentDate"))
.Append(mlContext.Transforms.NormalizeMeanVariance(outputColumnName: nameof(OHCL.PreviousClose)))
.Append(mlContext.Transforms.NormalizeMeanVariance(outputColumnName: nameof(OHCL.PreviousVolume)))
.Append(mlContext.Transforms.NormalizeMeanVariance(outputColumnName: nameof(OHCL.PreviousTrades)))
.Append(mlContext.Transforms.NormalizeMeanVariance(outputColumnName: nameof(OHCL.PreviousQuoeteAssetVolume)))
.Append(mlContext.Transforms.NormalizeMeanVariance(outputColumnName: nameof(OHCL.PreviousOpen)))
.Append(mlContext.Transforms.NormalizeMeanVariance(outputColumnName: nameof(OHCL.PreviousHigh)))
.Append(mlContext.Transforms.NormalizeMeanVariance(outputColumnName: nameof(OHCL.PreviousLow)))
.Append(mlContext.Transforms.Concatenate("Features", "CurrentDateE", nameof(OHCL.PreviousClose), nameof(OHCL.PreviousVolume), nameof(OHCL.PreviousTrades), nameof(OHCL.PreviousQuoeteAssetVolume), nameof(OHCL.PreviousOpen), nameof(OHCL.PreviousHigh), nameof(OHCL.PreviousLow)))
.Append(mlContext.Regression.Trainers.OnlineGradientDescent())
;
// STEP 3: Set the training algorithm, then create and config the modelBuilder - Selected Trainer (SDCA Regression algorithm)
var trainer = mlContext.Regression.Trainers.Sdca(labelColumnName: "Label", featureColumnName: "Features");
var trainingPipeline = pipeline.Append(trainer);
// 3. Train model
Console.WriteLine(dateBegin + " - Train...");
var model = trainingPipeline.Fit(trainingData);
Console.WriteLine("OK!");
OHCL last = getOHCL(dateBegin, loadData(DateTime.Now));
OHCL prediction = new OHCL();
prediction.CurrentDate = dateBegin;
prediction.PreviousClose = last.PreviousClose;
prediction.PreviousVolume = last.PreviousVolume;
prediction.PreviousTrades = last.PreviousTrades;
prediction.PreviousQuoeteAssetVolume = last.PreviousQuoeteAssetVolume;
prediction.PreviousOpen = last.PreviousOpen;
prediction.PreviousHigh = last.PreviousHigh;
prediction.PreviousLow = last.PreviousLow;
prediction.NextClose = 0;
var aux = mlContext.Model.CreatePredictionEngine <OHCL, OHCLPrediction>(model).Predict(prediction);
//Console.WriteLine(aux.NextClose);
if (last.PreviousClose < aux.NextClose)
{
longOrShort = "'L'";
}
if (last.PreviousClose > aux.NextClose)
{
longOrShort = "'S'";
}
lines += "['" + dateBegin.ToString("yyyy-MM-dd") + "'," + last.NextClose.ToString().Replace(",", ".") + "," + aux.NextClose.ToString().Replace(",", ".") + "," + longOrShort + "],";
dateBegin = dateBegin.AddDays(1);
}
}
catch (Exception ex)
{
Console.WriteLine(ex.Message + ex.StackTrace);
}
Console.WriteLine("press enter to exit...");
Console.ReadLine();
Environment.Exit(0);
}
