public static void InitializeForPrediction(List <DenseVector> dataSet, out CasesData trainingCases, out CasesData testCases,
int outputCount, int historyLength, float trainSetPercentage = 0.7f)
{
int inputCount = dataSet[0].Count() - outputCount;
if (inputCount < 1 || outputCount < 1)
{
throw new ArgumentException("Niepoprawne dane w plikach .csv");
}
if (trainSetPercentage <= 0 || trainSetPercentage > 1)
{
throw new ArgumentException("Train Set Percentage should be between 0 and 1");
}
List <DenseVector> trainSet = new List <DenseVector>();
List <DenseVector> testSet = new List <DenseVector>();
int atIndex = 0;
for (atIndex = 0; atIndex < dataSet.Count * trainSetPercentage; atIndex++)
{
trainSet.Add(dataSet[atIndex]);
}
if (trainSetPercentage == 1.0f) // special mode (testing on train set)
{
testSet = trainSet;
}
else
{
atIndex -= historyLength; // move back to have a "buffer" for historical data.
while (atIndex < dataSet.Count)
{
testSet.Add(dataSet[atIndex++]);
}
}
trainingCases = new CasesData(ProblemType.Regression,
trainSet.Select(x => x.CreateSubVector(0, inputCount))
.ToList(),
trainSet.Select(x => x.CreateSubVector(inputCount, outputCount))
.ToList(),
outputCount,
historyLength);
testCases = new CasesData(ProblemType.Regression,
testSet.Select(x => x.CreateSubVector(0, inputCount)).ToList(),
testSet.Select(x => x.CreateSubVector(inputCount, outputCount))
.ToList(),
outputCount,
historyLength);
}
public static void InitializeAllData(List <DenseVector> trainingData, List <DenseVector> testData, ProblemType problem,
out CasesData trainingCases, out CasesData testCases)
{
int input = testData.First().Count();
int output = trainingData.First().Count() - input;
if (input < 1 || output < 1)
{
throw new ArgumentException("Niepoprawne dane w plikach .csv");
}
trainingCases = new CasesData(problem,
trainingData.Select(x => x.CreateSubVector(0, input))
.ToList(),
trainingData.Select(x => x.CreateSubVector(input, output))
.ToList(),
output);
testCases = new CasesData(problem, testData, null,
(problem == ProblemType.Regression) ? output : trainingCases.ClassCount);
testCases.ClassIndexes = trainingCases.ClassIndexes;
}
