/// <summary>
/// This method reads the training data and calculates a variety of statistical data about each column.
/// </summary>
public static MetricStatistics UnderstandData(DataSetColumns dataSetColumns)
{
//Calculate and output the statistics for your pleasure
var stats = MetricStatistics.CalculateMetricStatisticsAsync(dataSetColumns.TimeSeries, DateTime.Now, TimeSpan.FromSeconds(1)).Result;
Console.WriteLine("Time Series Statistics:");
Console.WriteLine(stats.ToString());
return(stats);
}
public static void PopulateDataSet(DataSetColumns dataSetColumns)
{
var timeSeries = new List <double>();
using (var reader = new StreamReader(_dataPath2))
{
reader.ReadLine(); // headers
while (!reader.EndOfStream)
{
var line = reader.ReadLine();
var values = line.Split(',');
timeSeries.Add(double.Parse(values[1]));
}
}
dataSetColumns.TimeSeries = timeSeries.ToArray();
}
private static double[][] GetTrainingData(DataSetColumns dataSetColumns)
{
var length = dataSetColumns.TimeSeries.Length - FrameSize;
var retval = new double[length][];
// build array where first 24 values are input and the last value is the expected output
for (int i = 0; i < length; i++)
{
retval[i] = new double[FrameSize];
for (int j = 0; j < FrameSize; j++)
{
retval[i][j] = dataSetColumns.TimeSeries[i + j];
}
}
return(retval);
}
public static void ModelAndTrain(DataSetColumns dataSetColumns, double accuracy)
{
var trainData = GetTrainingData(dataSetColumns);
int numInput = InputSize; // number predictors
int numHidden = InputSize * 2;
int numOutput = 1; // regression
NeuralNetwork bestNeuralNetwork = null;
double bestTrainingAccuracy = 0.0;
for (int i = 0; i < 6; i++)
{
numHidden += numInput;
Console.WriteLine("Creating a " + numInput + "-" + numHidden +
"-" + numOutput + " neural network");
NeuralNetwork nn = new NeuralNetwork(numInput, numHidden, numOutput, true);
int maxEpochs = 10000;
double learnRate = 0.01;
Console.WriteLine("\nSetting maxEpochs = " + maxEpochs);
Console.WriteLine("Setting learnRate = " + learnRate.ToString("F2"));
Console.WriteLine("\nStarting training");
double[] weights = nn.Train(trainData, maxEpochs, learnRate);
Console.WriteLine("Done");
Console.WriteLine("\nFinal neural network model weights and biases:\n");
ShowVector(weights, 2, 10, true);
double trainAcc = nn.Accuracy(trainData, accuracy); // within standard deviation
if (bestNeuralNetwork == null || bestTrainingAccuracy < trainAcc)
{
bestTrainingAccuracy = trainAcc;
bestNeuralNetwork = nn;
}
Console.WriteLine("\nModel accuracy (+/- input accuracy) on training data = " +
trainAcc.ToString("F4"));
}
Console.WriteLine("\n========================= Best Model accuracy (+/- input accuracy) on training data = " +
bestTrainingAccuracy);
}
static void Main(string[] args)
{
try
{
// Get the data for initial
var datasetColumns = new DataSetColumns();
PopulateDataSet(datasetColumns);
// Learn about your data
var stats = UnderstandData(datasetColumns);
//Do Neural Networking
//ModelAndTrain(datasetColumns, stats.MedianAbsoluteDeviation);
}
catch (Exception e)
{
Console.WriteLine(e);
}
Console.WriteLine("Execution Complete -- Press Enter");
Console.ReadLine();
}
