private void TestBtn_Copy_Click(object sender, RoutedEventArgs e)
{
if (this.network == null)
{
return;
}
var classesCount = 0;
var attributesCount = 0;
var csvData = this.GetSetFromFile(ref classesCount, ref attributesCount, true);
if (csvData == null)
{
return;
}
var dataSet = csvData.NormalizedData;
var resX = new List <double>(dataSet.Count);
var resY = new List <double>(dataSet.Count);
var lowX = dataSet.Min(x => x.Item1[0]);
var highX = dataSet.Max(x => x.Item1[0]);
var lowY = dataSet.Min(x => x.Item1[1]);
var highY = dataSet.Max(x => x.Item1[1]);
var data = dataSet.Select(x => new { x = x.Item1[0], y = x.Item1[1], resCls = NetworkBase.GetClass(this.network.Predict(x.Item1)) });
this.DrawChart(
"Trained results",
data.GroupBy(x => x.resCls)
.Select(x => x.Select(y => new Tuple <double, double>(y.x, y.y))),
lowX,
highX,
lowY,
highY
);
SaveFileDialog saveFileDlg = new SaveFileDialog();
saveFileDlg.Filter = "csv files (*.csv)|*.txt|All files (*.*)|*.*";
var c = new CultureInfo("en-US");
if (saveFileDlg.ShowDialog() == true)
{
File.WriteAllLines
(
saveFileDlg.FileName,
data.Zip(csvData.RawData, (f, s) => $"{s.Item1[0].ToString(c)},{s.Item1[1].ToString(c)},{f.resCls + 1}")
);
}
}
private void TestBtnClick(object sender, RoutedEventArgs e)
{
if (this.network == null)
{
return;
}
int classesCount = 0, attributesCount = 0;
var testSet = this.GetSetFromFile(ref classesCount, ref attributesCount)?.NormalizedData;
if (testSet is null)
{
return;
}
var resX = new List <double>(testSet.Count);
var resY = new List <double>(testSet.Count);
var lowX = testSet.Min(x => x.Item1[0]);
var highX = testSet.Max(x => x.Item1[0]);
var lowY = 0.0;
var highY = 0.0;
if (networkType is ClassificationNetwork)
{
lowY = testSet.Min(x => x.Item1[1]);
highY = testSet.Max(x => x.Item1[1]);
}
else if (networkType is RegressionNetwork)
{
lowY = testSet.Min(x => x.Item2[0]);
highY = testSet.Max(x => x.Item2[0]);
}
if (networkType is ClassificationNetwork)
{
this.DrawChart
(
"Correct results",
testSet.GroupBy(x => x.Item2.IndexOf(1.0)).Select(g => g.Select(x => new Tuple <double, double>(x.Item1[0], x.Item1[1]))),
lowX,
highX,
lowY,
highY
);
}
else
{
this.DrawChart
(
"Correct results",
testSet.GroupBy(x => x.Item2.IndexOf(1.0)).Select(g => g.Select(x => new Tuple <double, double>(x.Item1[0], x.Item2[0]))),
lowX,
highX,
lowY,
highY
);
}
var data = testSet.Select(x => new { x = x.Item1[0], y = x.Item1[1], cls = NetworkBase.GetClass(x.Item2), resCls = NetworkBase.GetClass(this.network.Predict(x.Item1)) });
if (networkType is RegressionNetwork)
{
data = testSet.Select(x => new { x = x.Item1[0], y = this.network.Predict(x.Item1).First(), cls = NetworkBase.GetClass(x.Item2), resCls = NetworkBase.GetClass(this.network.Predict(x.Item2)) });
}
var acc = data.Count(x => x.cls == x.resCls) / (double)data.Count() * 100;
this.DrawChart(
"Trained results",
data.GroupBy(x => x.resCls)
.Select(x => x.Select(y => new Tuple <double, double>(y.x, y.y))),
lowX,
highX,
lowY,
highY
);
this.AccLbl.Content = acc.ToString();
}
