void FileTest(msbk57dlbe app)
{
string[] testData = File.ReadAllLines(@"tests.txt");
int TP = 0, TN = 0, FP = 0, FN = 0;
int goodPrediction = 0, wrongPrediction = 0;
foreach (string line in testData)
{
List <float> values = line.Split(';').Select(x => float.Parse(x, CultureInfo.InvariantCulture.NumberFormat)).ToList();
float pred = app.Prediction(values);
if (Math.Round(pred) != values[values.Count - 1])
{
Console.WriteLine("--- " + pred + "\t" + line);
wrongPrediction++;
if (pred == 1)
{
FP++;
}
else
{
FN++;
}
}
else
{
Console.WriteLine("+++ " + pred + "\t" + line);
goodPrediction++;
if (Math.Round(pred) == 1)
{
TP++;
}
else
{
TN++;
}
}
}
float accuracy = (float)(TP + TN) / (TP + FP + TN + FN);
float precision = (float)TP / (TP + FP);
float sensitivity = (float)TP / (TP + FN);
float F1 = 2 * (precision * sensitivity) / (precision + sensitivity);
Console.WriteLine(String.Format("True positive:\t{0}\nTrue negative:\t{1}\nFalse positive:\t{2}\nFalse negative:\t{3}", TP, TN, FP, FN));
Console.WriteLine(String.Format("Accuracy:\t{0}\nPrecision:\t{1}\nSensitivity:\t{2}\nF1 score:\t{3}", accuracy, precision, sensitivity, F1));
Console.WriteLine(String.Format("Good prediction:{0} ({1}%)", goodPrediction, 100f * goodPrediction / testData.Count()));
Console.WriteLine(String.Format("Wrong prediction:{0} ({1}%)", wrongPrediction, 100f * wrongPrediction / testData.Count()));
}
static void Main(string[] args)
{
DataSet.LoadMinMax(@"trains.txt");
//DataSet.GetTruePercents(@"trains.txt", DataSet.BinaryStartingIndex);
DataSet trainDS = new DataSet(@"trains.txt");
DataSet testDS = new DataSet(@"tests.txt");
msbk57dlbe app = new msbk57dlbe();
app.Train(trainDS);
app.Save(@"DeepNetworkTest2.model");
//msbk57dlbe app = new msbk57dlbe(@"DeepNetworkTest.model"); //vagy DeepNetwork.model
Console.WriteLine("Eval train:" + app.Evaluate(trainDS));
Console.WriteLine("Eval test:" + app.Evaluate(testDS));
new Program().FileTest(app);
Console.ReadKey();
}
