/// <summary>
/// Get all subsets of preferences that the skyline sampling algorithm will use to calculate its subset skylines. The subsets are calculated randomly.
/// </summary>
/// <remarks>
/// As long as the desired number of subsets (i.e., SubsetsCount) is not reached, new subsets are randomly
/// produced. When the number is reached, it is checked whether all preferences requested in the original skyline query
/// are contained in at least on subset. If this is not the case, one randomly chosen produced subset is removed
/// and the process continues to produce and remove subsets until both conditions are met.
/// </remarks>
/// <returns>
/// The produced subsets. Each stored integer is an index referring the zero-based position of a preference of
/// the original skyline query.
/// </returns>
/// <exception cref="Exception">Thrown on invocation of SkylineSamplingUtility.CheckValidityOfCountAndDimension.</exception>
public IEnumerable <CLRSafeHashSet <int> > GetSubsets()
{
SkylineSamplingUtility.CheckValidityOfCountAndDimension(SubsetsCount, SubsetDimension, AllPreferencesCount);
var subsetsReturn = new List <CLRSafeHashSet <int> >();
while (!IsSubsetProductionComplete(subsetsReturn))
{
if (subsetsReturn.Count >= SubsetsCount)
{
RemoveOneSubsetRandomly(subsetsReturn);
}
AddOneRandomSubsetNotYetContained(subsetsReturn);
}
return(subsetsReturn);
}
/// <summary>
/// Instantiates an object with the specified SkylineSamplingUtility as its Utility.
/// </summary>
/// <param name="utility">Used as Utility for the skyline sampling algorithm.</param>
internal SkylineSampling(SkylineSamplingUtility utility)
{
_utility = utility;
}
/// <summary>
/// Instantiates an object with the specified SkylineSamplingUtility as its Utility.
/// </summary>
/// <param name="utility">Used as Utility for the skyline sampling algorithm.</param>
internal SkylineSampling(SkylineSamplingUtility utility)
{
_utility = utility;
}
internal string MeasurePerformance(int iTrial, int iPreferenceIndex, ArrayList listPreferences,
ArrayList preferences,
SQLCommon parser, Stopwatch sw, List<long> reportDimensions, List<long> reportSkylineSize,
List<long> reportTimeTotal,
List<long> reportTimeAlgorithm, List<double> reportMinCorrelation, List<double> reportMaxCorrelation, double minCorrelation, double maxCorrelation,
List<double> reportCardinality, double cardinality,
string strSQL, string strPreferenceSet, string strTrial, List<long> reportNumberOfMoves, long numberOfMoves, List<long> reportNumberOfComparisons, long numberOfComparisons)
{
Dictionary<ClusterAnalysis, List<List<double>>> clusterAnalysis;
Dictionary<ClusterAnalysis, List<List<double>>> clusterAnalysisMedian;
Dictionary<ClusterAnalysis, Dictionary<BigInteger, List<double>>> clusterAnalysisTopBuckets;
Dictionary<ClusterAnalysis, Dictionary<BigInteger, List<double>>> clusterAnalysisMedianTopBuckets;
List<IEnumerable<CLRSafeHashSet<int>>> producedSubsets =
ProduceSubsets(preferences);
InitClusterAnalysisDataStructures(out clusterAnalysis);
InitClusterAnalysisDataStructures(out clusterAnalysisMedian);
InitClusterAnalysisTopBucketsDataStructures(
out clusterAnalysisTopBuckets);
InitClusterAnalysisTopBucketsDataStructures(
out clusterAnalysisMedianTopBuckets);
var entireSkylineDataTable = new DataTable();
if (ExcessiveTests)
{
entireSkylineDataTable =
parser.ParseAndExecutePrefSQL(Helper.ConnectionString, Helper.ProviderName,
strSQL);
}
else
{
entireSkylineDataTable =
parser.ParseAndExecutePrefSQL(Helper.ConnectionString, Helper.ProviderName,
strSQL+ " SAMPLE BY RANDOM_SUBSETS COUNT " + SubsetsCount +
" DIMENSION " + SubsetDimension);
}
List<long[]> entireDataTableSkylineValues =
parser.SkylineType.Strategy.SkylineValues;
int[] skylineAttributeColumns =
SkylineSamplingHelper.GetSkylineAttributeColumns(entireSkylineDataTable);
IReadOnlyDictionary<long, object[]> entireSkylineDatabase =
prefSQL.SQLSkyline.Helper.GetDatabaseAccessibleByUniqueId(
entireSkylineDataTable, 0);
IReadOnlyDictionary<long, object[]> entireSkylineNormalized =
prefSQL.SQLSkyline.Helper.GetDatabaseAccessibleByUniqueId(
entireSkylineDataTable, 0);
SkylineSamplingHelper.NormalizeColumns(entireSkylineNormalized,
skylineAttributeColumns);
DataTable entireDataTable;
IReadOnlyDictionary<long, object[]> entireDatabaseNormalized =
GetEntireDatabaseNormalized(parser, strSQL, skylineAttributeColumns,
out entireDataTable);
IReadOnlyDictionary<long, object[]> entireDatabase =
prefSQL.SQLSkyline.Helper.GetDatabaseAccessibleByUniqueId(
entireDataTable, 0);
IReadOnlyDictionary<BigInteger, List<IReadOnlyDictionary<long, object[]>>>
entireDatabaseBuckets =
prefSQL.Evaluation.ClusterAnalysis.GetBuckets(entireDatabaseNormalized,
skylineAttributeColumns);
IReadOnlyDictionary<int, List<IReadOnlyDictionary<long, object[]>>>
aggregatedEntireDatabaseBuckets =
prefSQL.Evaluation.ClusterAnalysis.GetAggregatedBuckets(entireDatabaseBuckets);
foreach (
KeyValuePair<BigInteger, List<IReadOnlyDictionary<long, object[]>>> s in
entireDatabaseBuckets.OrderByDescending(l => l.Value.Count)
.ThenBy(l => l.Key).Take(5))
{
double percent = (double) s.Value.Count / entireDatabaseNormalized.Count;
clusterAnalysisTopBuckets[ClusterAnalysis.EntireDb].Add(s.Key,
new List<double>());
for (var i = 0; i < producedSubsets.Count; i++)
// to enable generalized average calculation
{
clusterAnalysisTopBuckets[ClusterAnalysis.EntireDb][s.Key]
.Add(percent);
}
}
IReadOnlyDictionary<BigInteger, List<IReadOnlyDictionary<long, object[]>>>
entireSkylineBuckets =
prefSQL.Evaluation.ClusterAnalysis.GetBuckets(entireSkylineNormalized,
skylineAttributeColumns);
IReadOnlyDictionary<int, List<IReadOnlyDictionary<long, object[]>>>
aggregatedEntireSkylineBuckets =
prefSQL.Evaluation.ClusterAnalysis.GetAggregatedBuckets(entireSkylineBuckets);
FillTopBuckets(clusterAnalysisTopBuckets,
ClusterAnalysis.EntireSkyline, entireSkylineBuckets,
entireSkylineNormalized.Count, entireDatabaseNormalized.Count,
entireSkylineNormalized.Count);
foreach (
KeyValuePair<BigInteger, List<double>> bucket in
clusterAnalysisTopBuckets[ClusterAnalysis.EntireSkyline])
{
double percent =
clusterAnalysisTopBuckets[ClusterAnalysis.EntireSkyline][
bucket.Key][0];
for (var i = 1; i < producedSubsets.Count; i++)
// to enable generalized average calculation
{
clusterAnalysisTopBuckets[ClusterAnalysis.EntireSkyline][
bucket.Key].Add(percent);
}
}
var clusterAnalysisForMedian = new prefSQL.Evaluation.ClusterAnalysis(entireDatabaseNormalized,
skylineAttributeColumns);
IReadOnlyDictionary<BigInteger, List<IReadOnlyDictionary<long, object[]>>>
entireDatabaseMedianBuckets =
clusterAnalysisForMedian.GetBuckets(entireDatabaseNormalized,
skylineAttributeColumns, true);
IReadOnlyDictionary<int, List<IReadOnlyDictionary<long, object[]>>>
aggregatedEntireDatabaseMedianBuckets =
prefSQL.Evaluation.ClusterAnalysis.GetAggregatedBuckets(entireDatabaseMedianBuckets);
foreach (
KeyValuePair<BigInteger, List<IReadOnlyDictionary<long, object[]>>> s in
entireDatabaseMedianBuckets.OrderByDescending(l => l.Value.Count)
.ThenBy(l => l.Key).Take(5))
{
double percent = (double) s.Value.Count / entireDatabaseNormalized.Count;
clusterAnalysisMedianTopBuckets[ClusterAnalysis.EntireDb].Add(
s.Key,
new List<double>());
for (var i = 0; i < producedSubsets.Count; i++)
// to enable generalized average calculation
{
clusterAnalysisMedianTopBuckets[ClusterAnalysis.EntireDb][
s.Key]
.Add(percent);
}
}
IReadOnlyDictionary<BigInteger, List<IReadOnlyDictionary<long, object[]>>>
entireSkylineMedianBuckets =
clusterAnalysisForMedian.GetBuckets(entireSkylineNormalized,
skylineAttributeColumns, true);
IReadOnlyDictionary<int, List<IReadOnlyDictionary<long, object[]>>>
aggregatedEntireSkylineMedianBuckets =
prefSQL.Evaluation.ClusterAnalysis.GetAggregatedBuckets(entireSkylineMedianBuckets);
FillTopBuckets(clusterAnalysisMedianTopBuckets,
ClusterAnalysis.EntireSkyline, entireSkylineMedianBuckets,
entireSkylineNormalized.Count, entireDatabaseNormalized.Count,
entireSkylineNormalized.Count);
foreach (
KeyValuePair<BigInteger, List<double>> bucket in
clusterAnalysisMedianTopBuckets[
ClusterAnalysis.EntireSkyline])
{
double percent =
clusterAnalysisMedianTopBuckets[
ClusterAnalysis.EntireSkyline][bucket.Key][0];
for (var i = 1; i < producedSubsets.Count; i++)
// to enable generalized average calculation
{
clusterAnalysisMedianTopBuckets[
ClusterAnalysis.EntireSkyline][bucket.Key].Add(percent);
}
}
strSQL += " SAMPLE BY RANDOM_SUBSETS COUNT " + SubsetsCount +
" DIMENSION " + SubsetDimension;
string strQuery;
string operators;
int numberOfRecords;
string[] parameter;
PrefSQLModel prefSqlModel = parser.GetPrefSqlModelFromPreferenceSql(strSQL);
string ansiSql = parser.GetAnsiSqlFromPrefSqlModel(prefSqlModel);
prefSQL.SQLParser.Helper.DetermineParameters(ansiSql, out parameter,
out strQuery, out operators,
out numberOfRecords);
var subsetObjects = new List<long>();
var subsetTime = new List<long>();
var subsetTimeElapsed = new List<long>();
var setCoverageSecondRandom = new List<double>();
var setCoverageSample = new List<double>();
var setCoverageBestRank = new List<double>();
var setCoverageSumRank = new List<double>();
var representationErrorSecondRandom = new List<double>();
var representationErrorSample = new List<double>();
var representationErrorBestRank = new List<double>();
var representationErrorSumRank = new List<double>();
var representationErrorSumSecondRandom = new List<double>();
var representationErrorSumSample = new List<double>();
var representationErrorSumBestRank = new List<double>();
var representationErrorSumSumRank = new List<double>();
var dominatedObjectsCountSecondRandom = new List<double>();
var dominatedObjectsCountSample = new List<double>();
var dominatedObjectsCountBestRank = new List<double>();
var dominatedObjectsCountSumRank = new List<double>();
var dominatedObjectsOfBestObjectSecondRandom = new List<double>();
var dominatedObjectsOfBestObjectSample = new List<double>();
var dominatedObjectsOfBestObjectBestRank = new List<double>();
var dominatedObjectsOfBestObjectSumRank = new List<double>();
var subsetCount = 1;
foreach (IEnumerable<CLRSafeHashSet<int>> subset in producedSubsets)
{
Console.WriteLine(strPreferenceSet + " (" + subsetCount + " / " +
producedSubsets.Count + ")");
sw.Restart();
var subsetsProducer = new FixedSkylineSamplingSubsetsProducer(subset);
var utility = new SkylineSamplingUtility(subsetsProducer);
var skylineSample = new SkylineSampling(utility)
{
SubsetCount = prefSqlModel.SkylineSampleCount,
SubsetDimension = prefSqlModel.SkylineSampleDimension,
SelectedStrategy = parser.SkylineType
};
DataTable sampleSkylineDataTable = skylineSample.GetSkylineTable(strQuery,
operators);
sw.Stop();
subsetObjects.Add(sampleSkylineDataTable.Rows.Count);
subsetTime.Add(skylineSample.TimeMilliseconds);
subsetTimeElapsed.Add(sw.ElapsedMilliseconds);
IReadOnlyDictionary<long, object[]> sampleSkylineDatabase =
prefSQL.SQLSkyline.Helper.GetDatabaseAccessibleByUniqueId(
sampleSkylineDataTable, 0);
IReadOnlyDictionary<long, object[]> sampleSkylineNormalized =
prefSQL.SQLSkyline.Helper.GetDatabaseAccessibleByUniqueId(
sampleSkylineDataTable, 0);
SkylineSamplingHelper.NormalizeColumns(sampleSkylineNormalized,
skylineAttributeColumns);
IReadOnlyDictionary<long, object[]> secondRandomSampleDatabase = new Dictionary<long, object[]>();
IReadOnlyDictionary<long, object[]> secondRandomSampleNormalized = new Dictionary<long, object[]>();
IReadOnlyDictionary<long, object[]> entireSkylineDataTableBestRankDatabase = new Dictionary<long, object[]>();
IReadOnlyDictionary<long, object[]> entireSkylineDataTableSumRankDatabase = new Dictionary<long, object[]>();
IReadOnlyDictionary<long, object[]> entireSkylineDataTableBestRankNormalized = new Dictionary<long, object[]>();
IReadOnlyDictionary<long, object[]> entireSkylineDataTableSumRankNormalized = new Dictionary<long, object[]>();
if (ExcessiveTests)
{
secondRandomSampleDatabase =
SkylineSamplingHelper.GetRandomSample(entireSkylineDatabase,
sampleSkylineDataTable.Rows.Count);
var secondRandomSampleNormalizedToBeCreated = new Dictionary<long, object[]>();
foreach (KeyValuePair<long, object[]> k in secondRandomSampleDatabase)
{
var newValue = new object[k.Value.Length];
k.Value.CopyTo(newValue, 0);
secondRandomSampleNormalizedToBeCreated.Add(k.Key, newValue);
}
secondRandomSampleNormalized =
new ReadOnlyDictionary<long, object[]>(
secondRandomSampleNormalizedToBeCreated);
SkylineSamplingHelper.NormalizeColumns(secondRandomSampleNormalized,
skylineAttributeColumns);
entireSkylineDataTableBestRankNormalized =
GetEntireSkylineDataTableRankNormalized(entireSkylineDataTable.Copy(),
entireDataTableSkylineValues, skylineAttributeColumns,
sampleSkylineDataTable.Rows.Count, 1,
out entireSkylineDataTableBestRankDatabase);
entireSkylineDataTableSumRankNormalized =
GetEntireSkylineDataTableRankNormalized(entireSkylineDataTable.Copy(),
entireDataTableSkylineValues, skylineAttributeColumns,
sampleSkylineDataTable.Rows.Count, 2,
out entireSkylineDataTableSumRankDatabase);
IReadOnlyDictionary<long, object[]> baseRandomSampleNormalized =
SkylineSamplingHelper.GetRandomSample(entireSkylineNormalized,
sampleSkylineDataTable.Rows.Count);
double setCoverageCoveredBySecondRandomSample = SetCoverage.GetCoverage(
baseRandomSampleNormalized,
secondRandomSampleNormalized, skylineAttributeColumns) * 100.0;
double setCoverageCoveredBySkylineSample = SetCoverage.GetCoverage(
baseRandomSampleNormalized,
sampleSkylineNormalized, skylineAttributeColumns) * 100.0;
double setCoverageCoveredByEntireBestRank = SetCoverage.GetCoverage(
baseRandomSampleNormalized,
entireSkylineDataTableBestRankNormalized, skylineAttributeColumns) * 100.0;
double setCoverageCoveredByEntireSumRank = SetCoverage.GetCoverage(baseRandomSampleNormalized,
entireSkylineDataTableSumRankNormalized, skylineAttributeColumns) *
100.0;
setCoverageSecondRandom.Add(setCoverageCoveredBySecondRandomSample);
setCoverageSample.Add(setCoverageCoveredBySkylineSample);
setCoverageBestRank.Add(setCoverageCoveredByEntireBestRank);
setCoverageSumRank.Add(setCoverageCoveredByEntireSumRank);
Dictionary<long, double>.ValueCollection baseRepresentationErrorSecondRandomSample = SetCoverage
.GetRepresentationError(
GetReducedSkyline(entireSkylineNormalized, secondRandomSampleNormalized),
secondRandomSampleNormalized, skylineAttributeColumns);
Dictionary<long, double>.ValueCollection baseRepresentationErrorSkylineSample = SetCoverage.GetRepresentationError(
GetReducedSkyline(entireSkylineNormalized, sampleSkylineNormalized),
sampleSkylineNormalized, skylineAttributeColumns);
Dictionary<long, double>.ValueCollection baseRepresentationErrorEntireBestRank =
SetCoverage.GetRepresentationError(
GetReducedSkyline(entireSkylineNormalized,
entireSkylineDataTableBestRankNormalized),
entireSkylineDataTableBestRankNormalized, skylineAttributeColumns);
Dictionary<long, double>.ValueCollection baseRepresentationErrorEntireSumRank =
SetCoverage.GetRepresentationError(
GetReducedSkyline(entireSkylineNormalized,
entireSkylineDataTableSumRankNormalized),
entireSkylineDataTableSumRankNormalized, skylineAttributeColumns);
representationErrorSecondRandom.Add(baseRepresentationErrorSecondRandomSample.Max()*100.0);
representationErrorSample.Add(baseRepresentationErrorSkylineSample.Max() * 100.0);
representationErrorBestRank.Add(baseRepresentationErrorEntireBestRank.Max() * 100.0);
representationErrorSumRank.Add(baseRepresentationErrorEntireSumRank.Max() * 100.0);
representationErrorSumSecondRandom.Add(baseRepresentationErrorSecondRandomSample.Sum() * 100.0);
representationErrorSumSample.Add(baseRepresentationErrorSkylineSample.Sum() * 100.0);
representationErrorSumBestRank.Add(baseRepresentationErrorEntireBestRank.Sum() * 100.0);
representationErrorSumSumRank.Add(baseRepresentationErrorEntireSumRank.Sum() * 100.0);
var dominatedObjectsCountRandomSample =
new DominatedObjects(entireDatabase,
secondRandomSampleDatabase,
skylineAttributeColumns);
var dominatedObjectsCountSampleSkyline =
new DominatedObjects(entireDatabase, sampleSkylineDatabase,
skylineAttributeColumns);
var dominatedObjectsCountEntireSkylineBestRank =
new DominatedObjects(entireDatabase,
entireSkylineDataTableBestRankDatabase, skylineAttributeColumns);
var dominatedObjectsCountEntireSkylineSumRank =
new DominatedObjects(entireDatabase,
entireSkylineDataTableSumRankDatabase, skylineAttributeColumns);
dominatedObjectsCountSecondRandom.Add(
dominatedObjectsCountRandomSample.NumberOfDistinctDominatedObjects);
dominatedObjectsCountSample.Add(
dominatedObjectsCountSampleSkyline.NumberOfDistinctDominatedObjects);
dominatedObjectsCountBestRank.Add(
dominatedObjectsCountEntireSkylineBestRank.NumberOfDistinctDominatedObjects);
dominatedObjectsCountSumRank.Add(
dominatedObjectsCountEntireSkylineSumRank.NumberOfDistinctDominatedObjects);
dominatedObjectsOfBestObjectSecondRandom.Add(
dominatedObjectsCountRandomSample
.NumberOfObjectsDominatedByEachObjectOrderedByDescCount.First().Value);
dominatedObjectsOfBestObjectSample.Add(
dominatedObjectsCountSampleSkyline
.NumberOfObjectsDominatedByEachObjectOrderedByDescCount.First().Value);
dominatedObjectsOfBestObjectBestRank.Add(
dominatedObjectsCountEntireSkylineBestRank
.NumberOfObjectsDominatedByEachObjectOrderedByDescCount.First().Value);
dominatedObjectsOfBestObjectSumRank.Add(
dominatedObjectsCountEntireSkylineSumRank
.NumberOfObjectsDominatedByEachObjectOrderedByDescCount.First().Value);
IReadOnlyDictionary<BigInteger, List<IReadOnlyDictionary<long, object[]>>>
sampleBuckets =
prefSQL.Evaluation.ClusterAnalysis.GetBuckets(sampleSkylineNormalized,
skylineAttributeColumns);
IReadOnlyDictionary<int, List<IReadOnlyDictionary<long, object[]>>>
aggregatedSampleBuckets =
prefSQL.Evaluation.ClusterAnalysis.GetAggregatedBuckets(sampleBuckets);
IReadOnlyDictionary<BigInteger, List<IReadOnlyDictionary<long, object[]>>>
randomBuckets =
prefSQL.Evaluation.ClusterAnalysis.GetBuckets(secondRandomSampleNormalized,
skylineAttributeColumns);
IReadOnlyDictionary<int, List<IReadOnlyDictionary<long, object[]>>>
aggregatedRandomBuckets =
prefSQL.Evaluation.ClusterAnalysis.GetAggregatedBuckets(randomBuckets);
IReadOnlyDictionary<BigInteger, List<IReadOnlyDictionary<long, object[]>>>
bestRankBuckets =
prefSQL.Evaluation.ClusterAnalysis.GetBuckets(
entireSkylineDataTableBestRankNormalized,
skylineAttributeColumns);
IReadOnlyDictionary<int, List<IReadOnlyDictionary<long, object[]>>>
aggregatedBestRankBuckets =
prefSQL.Evaluation.ClusterAnalysis.GetAggregatedBuckets(
bestRankBuckets);
IReadOnlyDictionary<BigInteger, List<IReadOnlyDictionary<long, object[]>>>
sumRankBuckets =
prefSQL.Evaluation.ClusterAnalysis.GetBuckets(
entireSkylineDataTableSumRankNormalized,
skylineAttributeColumns);
IReadOnlyDictionary<int, List<IReadOnlyDictionary<long, object[]>>>
aggregatedSumRankBuckets =
prefSQL.Evaluation.ClusterAnalysis.GetAggregatedBuckets(
sumRankBuckets);
FillTopBuckets(clusterAnalysisTopBuckets,
ClusterAnalysis.SampleSkyline, sampleBuckets,
sampleSkylineNormalized.Count, entireDatabaseNormalized.Count,
entireSkylineNormalized.Count);
FillTopBuckets(clusterAnalysisTopBuckets,
ClusterAnalysis.RandomSkyline, randomBuckets,
secondRandomSampleNormalized.Count, entireDatabaseNormalized.Count,
entireSkylineNormalized.Count);
FillTopBuckets(clusterAnalysisTopBuckets,
ClusterAnalysis.BestRank, bestRankBuckets,
entireSkylineDataTableBestRankNormalized.Count,
entireDatabaseNormalized.Count, entireSkylineNormalized.Count);
FillTopBuckets(clusterAnalysisTopBuckets,
ClusterAnalysis.SumRank, sumRankBuckets,
entireSkylineDataTableSumRankNormalized.Count,
entireDatabaseNormalized.Count, entireSkylineNormalized.Count);
IReadOnlyDictionary<BigInteger, List<IReadOnlyDictionary<long, object[]>>>
sampleMedianBuckets =
clusterAnalysisForMedian.GetBuckets(sampleSkylineNormalized,
skylineAttributeColumns, true);
IReadOnlyDictionary<int, List<IReadOnlyDictionary<long, object[]>>>
aggregatedSampleMedianBuckets =
prefSQL.Evaluation.ClusterAnalysis.GetAggregatedBuckets(sampleMedianBuckets);
IReadOnlyDictionary<BigInteger, List<IReadOnlyDictionary<long, object[]>>>
randomMedianBuckets =
clusterAnalysisForMedian.GetBuckets(secondRandomSampleNormalized,
skylineAttributeColumns, true);
IReadOnlyDictionary<int, List<IReadOnlyDictionary<long, object[]>>>
aggregatedRandomMedianBuckets =
prefSQL.Evaluation.ClusterAnalysis.GetAggregatedBuckets(randomMedianBuckets);
IReadOnlyDictionary<BigInteger, List<IReadOnlyDictionary<long, object[]>>>
bestRankMedianBuckets =
clusterAnalysisForMedian.GetBuckets(
entireSkylineDataTableBestRankNormalized,
skylineAttributeColumns, true);
IReadOnlyDictionary<int, List<IReadOnlyDictionary<long, object[]>>>
aggregatedBestRankMedianBuckets =
prefSQL.Evaluation.ClusterAnalysis.GetAggregatedBuckets(
bestRankMedianBuckets);
IReadOnlyDictionary<BigInteger, List<IReadOnlyDictionary<long, object[]>>>
sumRankMedianBuckets =
clusterAnalysisForMedian.GetBuckets(
entireSkylineDataTableSumRankNormalized,
skylineAttributeColumns, true);
IReadOnlyDictionary<int, List<IReadOnlyDictionary<long, object[]>>>
aggregatedSumRankMedianBuckets =
prefSQL.Evaluation.ClusterAnalysis.GetAggregatedBuckets(
sumRankMedianBuckets);
FillTopBuckets(clusterAnalysisMedianTopBuckets,
ClusterAnalysis.SampleSkyline, sampleMedianBuckets,
sampleSkylineNormalized.Count, entireDatabaseNormalized.Count,
entireSkylineNormalized.Count);
FillTopBuckets(clusterAnalysisMedianTopBuckets,
ClusterAnalysis.RandomSkyline, randomMedianBuckets,
secondRandomSampleNormalized.Count, entireDatabaseNormalized.Count,
entireSkylineNormalized.Count);
FillTopBuckets(clusterAnalysisMedianTopBuckets,
ClusterAnalysis.BestRank, bestRankMedianBuckets,
entireSkylineDataTableBestRankNormalized.Count,
entireDatabaseNormalized.Count, entireSkylineNormalized.Count);
FillTopBuckets(clusterAnalysisMedianTopBuckets,
ClusterAnalysis.SumRank, sumRankMedianBuckets,
entireSkylineDataTableSumRankNormalized.Count,
entireDatabaseNormalized.Count, entireSkylineNormalized.Count);
var caEntireDbNew = new List<double>();
var caEntireSkylineNew = new List<double>();
var caSampleSkylineNew = new List<double>();
var caRandomSkylineNew = new List<double>();
var caBestRankNew = new List<double>();
var caSumRankNew = new List<double>();
for (var ii = 0; ii < skylineAttributeColumns.Length; ii++)
{
int entireSkyline = aggregatedEntireSkylineBuckets.ContainsKey(ii)
? aggregatedEntireSkylineBuckets[ii].Count
: 0;
int sampleSkyline = aggregatedSampleBuckets.ContainsKey(ii)
? aggregatedSampleBuckets[ii].Count
: 0;
int randomSkyline = aggregatedRandomBuckets.ContainsKey(ii)
? aggregatedRandomBuckets[ii].Count
: 0;
double entireSkylinePercent = (double)entireSkyline /
entireSkylineNormalized.Count;
double sampleSkylinePercent = (double) sampleSkyline /
sampleSkylineNormalized.Count;
double randomSkylinePercent = (double)randomSkyline /
secondRandomSampleNormalized.Count;
int entireDb = aggregatedEntireDatabaseBuckets.ContainsKey(ii)
? aggregatedEntireDatabaseBuckets[ii].Count
: 0;
double entireDbPercent = (double) entireDb /
entireDatabaseNormalized.Count;
int bestRank = aggregatedBestRankBuckets.ContainsKey(ii)
? aggregatedBestRankBuckets[ii].Count
: 0;
int sumRank = aggregatedSumRankBuckets.ContainsKey(ii)
? aggregatedSumRankBuckets[ii].Count
: 0;
double bestRankPercent = (double) bestRank /
entireSkylineDataTableBestRankNormalized.Count;
double sumRankPercent = (double) sumRank /
entireSkylineDataTableSumRankNormalized.Count;
caEntireDbNew.Add(entireDbPercent);
caEntireSkylineNew.Add(entireSkylinePercent);
caSampleSkylineNew.Add(sampleSkylinePercent);
caRandomSkylineNew.Add(randomSkylinePercent);
caBestRankNew.Add(bestRankPercent);
caSumRankNew.Add(sumRankPercent);
}
var caMedianEntireDbNew = new List<double>();
var caMedianEntireSkylineNew = new List<double>();
var caMedianSampleSkylineNew = new List<double>();
var caMedianRandomSkylineNew = new List<double>();
var caMedianBestRankNew = new List<double>();
var caMedianSumRankNew = new List<double>();
for (var ii = 0; ii < skylineAttributeColumns.Length; ii++)
{
int entireSkyline = aggregatedEntireSkylineMedianBuckets.ContainsKey(ii)
? aggregatedEntireSkylineMedianBuckets[ii].Count
: 0;
int sampleSkyline = aggregatedSampleMedianBuckets.ContainsKey(ii)
? aggregatedSampleMedianBuckets[ii].Count
: 0;
int randomSkyline = aggregatedRandomMedianBuckets.ContainsKey(ii)
? aggregatedRandomMedianBuckets[ii].Count
: 0;
double entireSkylinePercent = (double)entireSkyline /
entireSkylineNormalized.Count;
double sampleSkylinePercent = (double) sampleSkyline /
sampleSkylineNormalized.Count;
double randomSkylinePercent = (double)randomSkyline /
secondRandomSampleNormalized.Count;
int entireDb = aggregatedEntireDatabaseMedianBuckets.ContainsKey(ii)
? aggregatedEntireDatabaseMedianBuckets[ii].Count
: 0;
double entireDbPercent = (double) entireDb /
entireDatabaseNormalized.Count;
int bestRank = aggregatedBestRankMedianBuckets.ContainsKey(ii)
? aggregatedBestRankMedianBuckets[ii].Count
: 0;
int sumRank = aggregatedSumRankMedianBuckets.ContainsKey(ii)
? aggregatedSumRankMedianBuckets[ii].Count
: 0;
double bestRankPercent = (double) bestRank /
entireSkylineDataTableBestRankNormalized.Count;
double sumRankPercent = (double) sumRank /
entireSkylineDataTableSumRankNormalized.Count;
caMedianEntireDbNew.Add(entireDbPercent);
caMedianEntireSkylineNew.Add(entireSkylinePercent);
caMedianSampleSkylineNew.Add(sampleSkylinePercent);
caMedianRandomSkylineNew.Add(randomSkylinePercent);
caMedianBestRankNew.Add(bestRankPercent);
caMedianSumRankNew.Add(sumRankPercent);
}
clusterAnalysis[ClusterAnalysis.EntireDb].Add(caEntireDbNew);
clusterAnalysis[ClusterAnalysis.EntireSkyline].Add(
caEntireSkylineNew);
clusterAnalysis[ClusterAnalysis.SampleSkyline].Add(
caSampleSkylineNew);
clusterAnalysis[ClusterAnalysis.RandomSkyline].Add(
caRandomSkylineNew);
clusterAnalysis[ClusterAnalysis.BestRank].Add(
caBestRankNew);
clusterAnalysis[ClusterAnalysis.SumRank].Add(
caSumRankNew);
clusterAnalysisMedian[ClusterAnalysis.EntireDb].Add(
caMedianEntireDbNew);
clusterAnalysisMedian[ClusterAnalysis.EntireSkyline].Add(
caMedianEntireSkylineNew);
clusterAnalysisMedian[ClusterAnalysis.SampleSkyline].Add(
caMedianSampleSkylineNew);
clusterAnalysisMedian[ClusterAnalysis.RandomSkyline].Add(
caMedianRandomSkylineNew);
clusterAnalysisMedian[ClusterAnalysis.BestRank].Add(
caMedianBestRankNew);
clusterAnalysisMedian[ClusterAnalysis.SumRank].Add(
caMedianSumRankNew);
}
subsetCount++;
}
Dictionary<ClusterAnalysis, string> clusterAnalysisStrings =
GetClusterAnalysisStrings(skylineAttributeColumns, clusterAnalysis);
Dictionary<ClusterAnalysis, string> clusterAnalysisMedianStrings =
GetClusterAnalysisStrings(skylineAttributeColumns, clusterAnalysisMedian);
Dictionary<ClusterAnalysis, string> clusterAnalysisTopBucketsStrings =
GetClusterAnalysisTopBucketsStrings(clusterAnalysisTopBuckets, ExcessiveTests);
Dictionary<ClusterAnalysis, string> clusterAnalysisMedianTopBucketsStrings =
GetClusterAnalysisTopBucketsStrings(clusterAnalysisMedianTopBuckets, ExcessiveTests);
var time = (long) (subsetTime.Average() + .5);
var objects = (long) (subsetObjects.Average() + .5);
var elapsed = (long) (subsetTimeElapsed.Average() + .5);
Console.WriteLine("subsetTime");
foreach (var i in subsetTime)
{
Console.WriteLine(i);
}
Console.WriteLine("");
Console.WriteLine("subsetObjects");
foreach (var i in subsetObjects)
{
Console.WriteLine(i);
}
Console.WriteLine("");
reportDimensions.Add(preferences.Count);
reportSkylineSize.Add(objects);
reportTimeTotal.Add(elapsed);
reportTimeAlgorithm.Add(time);
reportMinCorrelation.Add(minCorrelation);
reportMaxCorrelation.Add(maxCorrelation);
reportCardinality.Add(cardinality);
var setCoverageSingle =
new Dictionary<SkylineTypesSingle, List<double>>
{
{SkylineTypesSingle.Random, setCoverageSecondRandom},
{SkylineTypesSingle.Sample, setCoverageSample},
{SkylineTypesSingle.BestRank, setCoverageBestRank},
{SkylineTypesSingle.SumRank, setCoverageSumRank}
};
var representationErrorSingle =
new Dictionary<SkylineTypesSingle, List<double>>
{
{SkylineTypesSingle.Random, representationErrorSecondRandom},
{SkylineTypesSingle.Sample, representationErrorSample},
{SkylineTypesSingle.BestRank, representationErrorBestRank},
{SkylineTypesSingle.SumRank, representationErrorSumRank}
};
var representationErrorSumSingle =
new Dictionary<SkylineTypesSingle, List<double>>
{
{SkylineTypesSingle.Random, representationErrorSumSecondRandom},
{SkylineTypesSingle.Sample, representationErrorSumSample},
{SkylineTypesSingle.BestRank, representationErrorSumBestRank},
{SkylineTypesSingle.SumRank, representationErrorSumSumRank}
};
var dominatedObjectsCountSingle =
new Dictionary<SkylineTypesSingle, List<double>>()
{
{
SkylineTypesSingle.Random,
dominatedObjectsCountSecondRandom
},
{
SkylineTypesSingle.Sample,
dominatedObjectsCountSample
},
{
SkylineTypesSingle.BestRank,
dominatedObjectsCountBestRank
},
{
SkylineTypesSingle.SumRank,
dominatedObjectsCountSumRank
}
};
var dominatedObjectsOfBestObjectSingle =
new Dictionary<SkylineTypesSingle, List<double>>
{
{
SkylineTypesSingle.Random,
dominatedObjectsOfBestObjectSecondRandom
},
{
SkylineTypesSingle.Sample,
dominatedObjectsOfBestObjectSample
},
{
SkylineTypesSingle.BestRank,
dominatedObjectsOfBestObjectBestRank
},
{
SkylineTypesSingle.SumRank,
dominatedObjectsOfBestObjectSumRank
}
};
AddToReports(_reportsLong, subsetObjects, subsetTime,
_reportsDouble);
if (ExcessiveTests)
{
AddToSetCoverage(_setCoverage, setCoverageSingle);
AddToSetCoverage(_representationError,
representationErrorSingle);
AddToSetCoverage(_representationErrorSum,
representationErrorSumSingle);
AddToSetCoverage(_dominatedObjectsCount,
dominatedObjectsCountSingle);
AddToSetCoverage(_dominatedObjectsOfBestObject,
dominatedObjectsOfBestObjectSingle);
}
string strLine = FormatLineString(strPreferenceSet, strTrial,
preferences.Count, objects,
elapsed, time, subsetTime.Min(), subsetTime.Max(),
MyMathematic.GetSampleVariance(subsetTime),
MyMathematic.GetSampleStdDeviation(subsetTime),
Mathematic.Median(subsetTime), Mathematic.LowerQuartile(subsetTime),
Mathematic.UpperQuartile(subsetTime), subsetObjects.Min(),
subsetObjects.Max(), MyMathematic.GetSampleVariance(subsetObjects),
MyMathematic.GetSampleStdDeviation(subsetObjects),
Mathematic.Median(subsetObjects), Mathematic.LowerQuartile(subsetObjects),
Mathematic.UpperQuartile(subsetObjects),
setCoverageSingle, representationErrorSingle,
representationErrorSumSingle, dominatedObjectsCountSingle,
dominatedObjectsOfBestObjectSingle,
clusterAnalysisStrings, clusterAnalysisMedianStrings,
clusterAnalysisTopBucketsStrings, clusterAnalysisMedianTopBucketsStrings,
minCorrelation, maxCorrelation,
cardinality, numberOfMoves, numberOfComparisons);
return strLine;
}
