/// <summary>
/// 计算峰值
/// </summary>
/// <param name="N">回览次数</param>
/// <param name="K">成功次数</param>
/// <param name="p">正常概率</param>
public override List <ChanceClass> getChances(BaseCollection sc, ExpectData ed)
{
List <ChanceClass> ret = new List <ChanceClass>();
MLDataFactory pkdls = new MLDataFactory();// this.LastUseData());
//Dictionary<int, Dictionary<int, int>> res = pkdls.getAllShiftAndColMaxProbList(this.ReviewExpectCnt- this.InputMinTimes-1, this.InputMinTimes,true);
List <MLFeature <int> > features = pkdls.getAllSpecRowRoundFeatures(this.LastUseData().Count - 1, this.InputMinTimes, 0);
Dictionary <string, int> AllCodes = new Dictionary <string, int>();
for (int i = 0; i < features.Count; i++)
{
MLFeature <int> feature = features[i];
MaxEnt me = new MaxEnt();
if (MaxEnt.FeatureSummary == null)
{
me.OnLoadLocalFile = LoadLocalTrainData;
me.LoadSummary();
}
me.FillStructBySummary(i);
MLInstance <int, int> instance = new MLInstance <int, int>(feature);
int label = me.Classify(instance);
if (label < 0)
{
continue;
}
string strAllCode = string.Format("{0}/{1}", (i + 1) % 10, label);
AllCodes.Add(strAllCode, 1);
}
////foreach (int key in res.Keys)
//// AllCodes.Add(string.Format("{0}/{1}",key,string.Join("",res[key].Keys.ToArray())),1);
if (true)
{
ChanceClass cc = new ChanceClass();
cc.SignExpectNo = ed.Expect;
cc.ChanceType = 2;
cc.InputTimes = 1;
cc.strInputTimes = string.Join("_", new string[2] {
"1", "1"
});
//cc.AllowMaxHoldTimeCnt = this.AllowMaxHoldTimeCnt;
cc.InputExpect = ed;
cc.StragId = this.GUID;
//cc.MinWinRate = this.CommSetting.GetGlobalSetting().DefaultHoldAmtSerials.MaxRates[cc.ChipCount - 1];
cc.IsTracer = 0;
cc.HoldTimeCnt = 1;
cc.Odds = this.CommSetting.GetGlobalSetting().Odds;
cc.ExpectCode = ed.Expect;
cc.ChanceCode = string.Join("+", AllCodes.Keys.ToArray());
cc.ChipCount = ChanceClass.getChipsByCode(cc.ChanceCode);
cc.CreateTime = DateTime.Now;
cc.UpdateTime = DateTime.Now;
cc.AllowMaxHoldTimeCnt = 1;
cc.Closed = false;
ret.Add(cc);
}
return(ret);
}
public MLInstance <int, int> getSpecRowRoundLabelFeatures(long rowid, int col, int Deep, int AllowUseShift)
{
MLInstance <int, int> ret = new MLInstance <int, int>();
if ((rowid - 1) - Deep < 0)//rowid 至少要大于等于Deep;
{
return(ret);
}
ret.Feature = getSpecRowRoundFeatures(rowid - 1, col, Deep, AllowUseShift);
ret.Label = int.Parse(Data[(int)rowid].ValueList[col]);
return(ret);
}
MLInstance <int, int> getSpecLengthMatchTimesProb(int col, int Deep, int AllowUseShift)
{
MLInstance <int, int> ret = new MLInstance <int, int>();
return(ret);
}
/// <summary>
/// Executes a run of processing the stock data
/// </summary>
/// <param name="session">The session configuration</param>
public static void Run(StockSession session)
{
// Create the stock processor
StockProcessor processor = new StockProcessor(session);
// Get the sizes of the features and labels
int dstIdx = 0;
int numDataPoints = 0;
int numFeatures, numLabels;
float[,] tmpDst = new float[1, 1024];
int[,] tmpLabels = new int[1, 1];
var tmpDataPoint = processor.DerivedData.First().Value[0];
tmpDataPoint.Load(session);
PopulateData(tmpDataPoint, tmpDst, tmpLabels, ref dstIdx, out numFeatures, out numLabels);
// Determine the total number of data points
List <string> symbols = new List <string>()
{
"GNTX"
};
foreach (var s in symbols)
{
List <StockDataSet <StockDataSink> > sources;
if (!processor.DerivedData.TryGetValue(s, out sources))
{
continue;
}
for (int i = 0; i < sources.Count; i++)
{
sources[i].Load(session);
numDataPoints += sources[i].Count;
}
}
// Allocate the feature and label arrays
float[,] features = new float[numDataPoints, numFeatures];
int[,] labels = new int[numDataPoints, numLabels];
// Load the data
dstIdx = 0;
foreach (var s in symbols)
{
List <StockDataSet <StockDataSink> > sources;
if (!processor.DerivedData.TryGetValue(s, out sources))
{
continue;
}
// Create a table of each data point in the specified range
for (int i = 0; i < sources.Count; i++)
{
sources[i].Load(session);
PopulateData(sources[i], features, labels, ref dstIdx, out numFeatures, out numLabels);
}
}
// Create the Machine Learning instance
MLInstance ml = new MLInstance();
ml.BuildFullyConnectedGraph(new int[] { numFeatures, numFeatures, numLabels });
ml.PrepareData(features, labels);
ml.Train();
}
