public static List <TotalScenario> GetTotalScenarios(List <Scenario> scenarios)
{
var totalScenarios = new List <TotalScenario>();
var ks = scenarios.Select(n => n.K).Distinct();
var algorithms = scenarios.Select(n => n.Algorithm).Distinct();
var trainingSizes = scenarios.Select(n => n.TrainingSize).Distinct();
foreach (var k in ks)
{
foreach (var algorithm in algorithms)
{
foreach (var trainingSize in trainingSizes)
{
var filtered = scenarios.Where(n => n.K == k && n.Algorithm == algorithm && n.TrainingSize == trainingSize);
var avgCorrect = filtered.Avg(n => n.Correct);
var avgAccuracy = filtered.Avg(n => n.Accuracy);
var stdCorrect = filtered.Std(n => n.Correct, avgCorrect);
var stdAccuracy = filtered.Std(n => n.Accuracy, avgAccuracy);
var scenario = new TotalScenario()
{
_trainingSize = trainingSize,
_k = k,
_algorithm = algorithm,
_avgCorrect = avgCorrect,
_stdCorrect = stdCorrect,
_avgAccuracy = avgAccuracy,
_stdAccuracy = stdAccuracy
};
totalScenarios.Add(scenario);
}
}
}
return(totalScenarios);
}
static void Main(string[] args)
{
var scenarios = new List <Scenario>();
Loader loader = null;
List <Point> dataset = null;
var ks = new List <int>();
var trainingSizes = new List <int>();
int iterations = 0;
string currentArg = null;
Console.WriteLine("Reading data...");
try
{
for (int i = 0; i < args.Length; i++)
{
if (i == 0)
{
loader = new Loader(args[i]);
dataset = loader.LoadPoints();
}
else if (i == 1)
{
if (!int.TryParse(args[i], out iterations) || iterations <= 0 || (iterations % 4 != 0 && iterations != 1))
{
Console.WriteLine("Invalid number of iterations");
Console.ReadKey();
return;
}
}
else if (args[i] == "t")
{
currentArg = args[i];
}
else if (args[i] == "k")
{
currentArg = args[i];
}
else if (currentArg == "t")
{
int t;
if (!int.TryParse(args[i], out t) || t >= dataset.Count - 1 || t <= 0)
{
Console.WriteLine("Invalid training size: {0}", t);
Console.ReadKey();
return;
}
trainingSizes.Add(t);
}
else if (currentArg == "k")
{
int k;
if (!int.TryParse(args[i], out k) || k <= 0)
{
Console.WriteLine("Invalid k: {0}", k);
Console.ReadKey();
return;
}
ks.Add(k);
}
}
}
catch (IOException e)
{
Console.WriteLine("Invalid path");
Console.ReadKey();
return;
}
if (ks.Count == 0 || trainingSizes.Count == 0)
{
Console.WriteLine("Not enough args, usage: ./program.exe dataset_path t 100 50 20 k 5 6 7");
Console.ReadKey();
return;
}
foreach (var size in trainingSizes)
{
foreach (var k in ks)
{
Console.WriteLine("Batch: size - {0}, k - {1}", size, k);
if (iterations > 1)
{
for (int j = 0; j < iterations; j++)
{
var p = Randomize(dataset);
var s = new Scenario(new MeanKnn(), loader, size, k);
s.Execute(p.Select(pt => new Point()
{
X = pt.X,
Y = pt.Y,
DistanceToOrigin = pt.DistanceToOrigin,
TrueLabel = pt.TrueLabel
}).ToList());
Console.WriteLine("Mean :" + s.Accuracy);
scenarios.Add(s);
s = new Scenario(new SimpleSolver(), loader, size, k);
s.Execute(p.Select(pt => new Point()
{
X = pt.X,
Y = pt.Y,
DistanceToOrigin = pt.DistanceToOrigin,
TrueLabel = pt.TrueLabel
}).ToList());
Console.WriteLine("Simple :" + s.Accuracy);
scenarios.Add(s);
}
}
else
{
var s = new Scenario(new MeanKnn(), loader, size, k);
s.Execute(dataset.Select(pt => new Point()
{
X = pt.X,
Y = pt.Y,
DistanceToOrigin = pt.DistanceToOrigin,
TrueLabel = pt.TrueLabel
}).ToList());
scenarios.Add(s);
s = new Scenario(new SimpleSolver(), loader, size, k);
s.Execute(dataset.Select(pt => new Point()
{
X = pt.X,
Y = pt.Y,
DistanceToOrigin = pt.DistanceToOrigin,
TrueLabel = pt.TrueLabel
}).ToList());
scenarios.Add(s);
}
}
}
Console.WriteLine("Saving results...");
var date = DateTime.Now;
var resultWriter = new ResultWriter();
resultWriter.WriteResults(scenarios, date);
resultWriter.WriteResults(TotalScenario.GetTotalScenarios(scenarios), date);
Console.WriteLine("Done.");
Console.ReadKey();
}
