public void ClusteringTest()
{
List <ClusteredData> datanew = new List <ClusteredData>();
EuclidianDistance distance = new EuclidianDistance();
K_Means k_means = new K_Means(2, 0.1, distance);
ClusterAnalysis clusterAnalysis = new ClusterAnalysis(k_means);
datanew.Add(new ClusteredData(0, 0));
datanew.Add(new ClusteredData(0, 20));
datanew.Add(new ClusteredData(20, 0));
datanew.Add(new ClusteredData(20, 20));
var list = clusterAnalysis.Clustering(datanew);
Assert.AreEqual(2, list.Count);
var flag = from value in k_means.FinishesCentroids.Values
where (value.X == 0 && value.Y == 0) ||
(value.X == 0 && value.Y == 20) ||
(value.X == 20 && value.Y == 20) ||
(value.X == 20 && value.Y == 0) ||
(value.X == 10 && value.Y == 20) ||
(value.X == 0 && value.Y == 10) ||
(value.X == 20 && value.Y == 10) ||
(value.X == 0 && value.Y == 10)
select value;
Assert.AreEqual(0.1, k_means.CoefficientTaboo);
Assert.AreEqual(2, k_means.CountOfClusters);
Assert.IsNotNull(flag);
}
/// <summary>
/// Converts the <see cref="sourceValue" /> parameter to the <see cref="destinationType" /> parameter using <see cref="formatProvider"
/// /> and <see cref="ignoreCase" />
/// </summary>
/// <param name="sourceValue">the value to convert into an instance of <see cref="ClusterAnalysis" />.</param>
/// <returns>
/// an instance of <see cref="ClusterAnalysis" />, or <c>null</c> if there is no suitable conversion.
/// </returns>
public static object ConvertFrom(dynamic sourceValue)
{
if (null == sourceValue)
{
return(null);
}
try
{
ClusterAnalysis.FromJsonString(typeof(string) == sourceValue.GetType() ? sourceValue : sourceValue.ToJsonString());
}
catch
{
// Unable to use JSON pattern
}
try
{
return(new ClusterAnalysis
{
VmEfficiencyMap = sourceValue.VmEfficiencyMap,
});
}
catch
{
}
return(null);
}
public void ClusterAnalysisMain(int countCluster, double[] weight, bool normalization)
{
ClusterAnalysis clusterAnalysis = new ClusterAnalysis();
double[,] setForm = new double[30, 3] // Входное множество образов для кластеризации
{
{ 15, 6, 15 },
{ 1, 19, 0 },
{ -8, -1, 4 },
{ 13, 19, 15 },
{ 15, 17, -14 },
{ -3, 9, -35 },
{ 12, 4, 16 },
{ 8, 14, 9 },
{ -6, 0, 5 },
{ 11, 17, 10 },
{ 12, 17, -10 },
{ -1, 10, -25 },
{ 18, 17, -11 },
{ -4, 9, -31 },
{ 19, 4, 13 },
{ 8, 14, 10 },
{ -6, -5, 1 },
{ 20, 20, 20 },
{ 7, 16, -17 },
{ -1, 7, -26 },
{ 15, 1, 10 },
{ 0, 11, 8 },
{ -8, -1, 5 },
{ 10, 10, 10 },
{ 12, 15, -10 },
{ -4, 5, -27 },
{ -7, -1, 4 },
{ 3, 17, 11 },
{ 0, 1, 3 },
{ 7, 2, 0 }
};
// Применение нормализации
if (normalization)
{
setForm = GetNormineSetForm(setForm);
}
ClusterCalculationByTwoMethods(setForm, countCluster, weight);
}
static void Main(string[] args)
{
List <ClusteredData> datanew = new List <ClusteredData>();
EuclidianDistance distance = new EuclidianDistance();
K_Means k_means = new K_Means(2, 0.1, distance);
ClusterAnalysis clusterAnalysis = new ClusterAnalysis(k_means);
datanew.Add(new ClusteredData(0, 0));
datanew.Add(new ClusteredData(0, 20));
datanew.Add(new ClusteredData(20, 0));
datanew.Add(new ClusteredData(20, 20));
var list = clusterAnalysis.Clustering(datanew);
/*
* List<ClusteredData> datanew = new List<ClusteredData>();
* List<K_Means> k_meansList = new List<K_Means>();
* List<List<Cluster>> clustersList = new List<List<Cluster>>();
* System.Diagnostics.Stopwatch myStopwatch = new System.Diagnostics.Stopwatch();
* EuclidianDistance distance = new EuclidianDistance();
*
* datanew.Add(new ClusteredData(0,0));
* datanew.Add(new ClusteredData(0, 20));
* datanew.Add(new ClusteredData(20, 0));
* datanew.Add(new ClusteredData(20, 20));
*
* var ewrt = datanew.Select(Er => Er.X);
* var ety = ewrt.Average();
*
*
*
* var e = datanew.Average(er => er.X);
*
*
* myStopwatch.Start(); //запуск
* for (var i = 0; i < 1000; i++)
* {
* K_Means k_means = new K_Means(2, 0.1, distance);
* clustersList.Add(k_means.Clustering(datanew));
* k_meansList.Add(k_means);
* Thread.Sleep(10);
*
* var flag = k_means.FinishesCentroids.ContainsValue(new Centroid(0, 0));
* var flag1 = k_means.FinishesCentroids.ContainsValue(new Centroid(0, 20));
* var flag2 = k_means.FinishesCentroids.ContainsValue(new Centroid(20, 20));
* var flag3 = k_means.FinishesCentroids.ContainsValue(new Centroid(20, 0));
* }
* myStopwatch.Stop();
*
* var fulltime = myStopwatch.ElapsedMilliseconds;
* double Onetime = myStopwatch.ElapsedMilliseconds/1000.0;
*
* int a = 0;
* int b = 0;
* int c = 0;
* int d = 0;
*
* foreach (var list in clustersList)
* {
* a += list.Count(q => q.Data.Count == 3);
* b += list.Count(q => q.Data.Count == 2);
* c += list.Count(q => q.Data.Count == 1);
* d += list.Count(q => q.Data.Count == 0);
* }
* b = b / 2;
* var ads = new Cluster(1, new Centroid(1, 10));
* //ads.IntraClusterDistance();
*
* //// Поиск объектов класса Cluster в списке optimumClusters с максимальным внутрикластерным расcтоянием
* //var nonOptimumClusters = from list in clustersList // определяем каждый объект из clusters как cluster
* // where
* // select cluster.IntraClusterDistance();// выбираем объект
*
*/
}
private void TestForClusterAnalysis()
{
var common = new SQLCommon
{
SkylineType =
new SkylineBNL()
{
Provider = Helper.ProviderName, ConnectionString = Helper.ConnectionString
},
ShowInternalAttributes = true
};
DbProviderFactory factory = DbProviderFactories.GetFactory(Helper.ProviderName);
// use the factory object to create Data access objects.
DbConnection connection = factory.CreateConnection();
// will return the connection object (i.e. SqlConnection ...)
connection.ConnectionString = Helper.ConnectionString;
var dt = new DataTable();
connection.Open();
DbDataAdapter dap = factory.CreateDataAdapter();
DbCommand selectCommand = connection.CreateCommand();
selectCommand.CommandTimeout = 0; //infinite timeout
string strQuery;
string operators;
int numberOfRecords;
string[] parameter;
string ansiSql =
common.GetAnsiSqlFromPrefSqlModel(common.GetPrefSqlModelFromPreferenceSql(_entireSkylineSql));
prefSQL.SQLParser.Helper.DetermineParameters(ansiSql, out parameter, out strQuery, out operators,
out numberOfRecords);
selectCommand.CommandText = strQuery;
dap.SelectCommand = selectCommand;
dt = new DataTable();
dap.Fill(dt);
DataTable entireSkylineDataTable = common.ParseAndExecutePrefSQL(Helper.ConnectionString,
Helper.ProviderName,
_entireSkylineSql);
int[] skylineAttributeColumns = SkylineSamplingHelper.GetSkylineAttributeColumns(entireSkylineDataTable);
IReadOnlyDictionary <long, object[]> entireSkylineNormalized =
prefSQL.SQLSkyline.Helper.GetDatabaseAccessibleByUniqueId(entireSkylineDataTable, 0);
SkylineSamplingHelper.NormalizeColumns(entireSkylineNormalized, skylineAttributeColumns);
DataTable sampleSkylineDataTable = common.ParseAndExecutePrefSQL(Helper.ConnectionString,
Helper.ProviderName,
_skylineSampleSql);
IReadOnlyDictionary <long, object[]> sampleSkylineNormalized =
prefSQL.SQLSkyline.Helper.GetDatabaseAccessibleByUniqueId(sampleSkylineDataTable, 0);
SkylineSamplingHelper.NormalizeColumns(sampleSkylineNormalized, skylineAttributeColumns);
for (var i = 0; i < skylineAttributeColumns.Length; i++)
{
dt.Columns.RemoveAt(0);
}
IReadOnlyDictionary <long, object[]> full = prefSQL.SQLSkyline.Helper.GetDatabaseAccessibleByUniqueId(dt, 0);
SkylineSamplingHelper.NormalizeColumns(full, skylineAttributeColumns);
ClusterAnalysis.CalcMedians(full, skylineAttributeColumns);
IReadOnlyDictionary <BigInteger, List <IReadOnlyDictionary <long, object[]> > > entireBuckets =
ClusterAnalysis.GetBuckets(entireSkylineNormalized, skylineAttributeColumns);
IReadOnlyDictionary <BigInteger, List <IReadOnlyDictionary <long, object[]> > > sampleBuckets =
ClusterAnalysis.GetBuckets(sampleSkylineNormalized, skylineAttributeColumns);
//IReadOnlyDictionary<int, List<IReadOnlyDictionary<long, object[]>>> aggregatedEntireBuckets =
// ClusterAnalysis.GetAggregatedBuckets(entireBuckets);
//IReadOnlyDictionary<int, List<IReadOnlyDictionary<long, object[]>>> aggregatedSampleBuckets =
// ClusterAnalysis.GetAggregatedBuckets(sampleBuckets);
IReadOnlyDictionary <BigInteger, List <IReadOnlyDictionary <long, object[]> > > fullB =
ClusterAnalysis.GetBuckets(full, skylineAttributeColumns);
// IReadOnlyDictionary<int, List<IReadOnlyDictionary<long, object[]>>> aFullB =
// ClusterAnalysis.GetAggregatedBuckets(fullB);
IOrderedEnumerable <KeyValuePair <BigInteger, List <IReadOnlyDictionary <long, object[]> > > > sorted = fullB.OrderBy(l => l.Value.Count)
.ThenBy(l => l.Key);
int len = Convert.ToInt32(Math.Pow(2, skylineAttributeColumns.Length));
//for (var i = 0; i < len; i++)
foreach (KeyValuePair <BigInteger, List <IReadOnlyDictionary <long, object[]> > > s in sorted)
{
BigInteger i = s.Key;
int entire = entireBuckets.ContainsKey(i) ? entireBuckets[i].Count : 0;
int sample = sampleBuckets.ContainsKey(i) ? sampleBuckets[i].Count : 0;
double entirePercent = (double)entire / entireSkylineNormalized.Count;
double samplePercent = (double)sample / sampleSkylineNormalized.Count;
int fullX = fullB.ContainsKey(i) ? fullB[i].Count : 0;
double fullP = (double)fullX / full.Count;
Console.WriteLine("-- {0,5} -- {5,6} ({6,7:P2} %) -- {1,6} ({3,7:P2} %) -- {2,6} ({4,7:P2} %)", i,
entire, sample, entirePercent,
samplePercent, fullX, fullP);
}
Console.WriteLine();
Console.WriteLine("{0} - {1} - {2}", entireSkylineNormalized.Count, sampleSkylineNormalized.Count,
full.Count);
}
