/// <summary>
/// Starts the k-NearestNeighbor Algorithm using the EuclideanDistance Method.
/// </summary>
/// <param name="k">The number of Chunks to use.</param>
public void Start_kNN(int k, int kfold, CalculationMethod method = CalculationMethod.EuclideanDistance)
{
DistanceCalculator calculationMethod = new EuclideanDistanceStrategy();
if (method == CalculationMethod.ManhattanDistance)
{
calculationMethod = new ManhattanDistanceStrategy();
}
else if (method == CalculationMethod.CorrectedEuclideanDistance)
{
calculationMethod = new CorrectedEuclideanStrategy();
}
Console.WriteLine("Calculating with method: {0} ...", method.ToString());
#region Calculate Chunk Size
SplitDataIntoChunks(kfold);
#endregion
#region Heuristic
// The Result
int[,] confusionMatrix = new int[10, 10];
Stopwatch stopWatch = new Stopwatch();
stopWatch.Start();
foreach (var testingChunk in _Chunks)
{
foreach (var unknownWine in testingChunk.Value) // For each wine, which is not part of the current Chunk
{
_NearestNeighbor = new SortedList <float, int>(new DuplicateKeyComparer <float>());
// Get NearestNeighbor
foreach (var nnChunk in _Chunks.Where(x => testingChunk.Key != x.Key))
{
foreach (var knownWine in nnChunk.Value)
{
float currDistance = calculationMethod.GetDistance(knownWine, unknownWine);
_NearestNeighbor.Add(currDistance, knownWine.Quality);
}
}
// Get most occuring quality
var w = _NearestNeighbor.Values.Take(k);
var dict = w.ToLookup(x => x);
var maxCount = dict.Max(x => x.Count());
int mostQuality = dict.Where(x => x.Count() == maxCount).Select(x => x.Key).FirstOrDefault();
// Add the current guess to the result
confusionMatrix[unknownWine.Quality, mostQuality]++;
}
}
stopWatch.Stop();
TimeSpan ts = stopWatch.Elapsed;
string elapsedTime = String.Format("{0} Minutes, {1} Seconds, {2} ms", ts.Minutes, ts.Seconds, ts.Milliseconds);
PrintResult(confusionMatrix, elapsedTime);
#endregion
}