public ApproximationSpace(
DecisionSystem infoSys,
Func<double?[], double?[], bool> indiscerRel,
Func<IEnumerable<double?[]>, IEnumerable<double?[]>, double> roughInclusion)
{
IS = infoSys;
IndRelation = indiscerRel;
RoughIncl = roughInclusion;
}
public static void TestReduct()
{
// Prepare test data
var infSys = new DecisionSystem
{ // KhuPN, Hong - "Datamining based on Rough set theory" pg. 16
Universe = new double?[][]
{ // id 1 2 3 4 5
new double?[]{ 1, 2, 1, 0, 0, 1},
new double?[]{ 2, 0, 2, 1, 2, 1},
new double?[]{ 3, 1, 2, 1, 1, 2},
new double?[]{ 4, 1, 0, 0, 1, 1},
new double?[]{ 5, 2, 1, 2, 0, 0},
new double?[]{ 6, 1, 0, 1, 1, 2},
new double?[]{ 7, 0, 2, 1, 0, 1},
new double?[]{ 8, 0, 1, 2, 1, 2}
},
ConditionAttributes = new[] { 1, 2, 3, 4},
DecisionAttributes = new[] { 5 },
DecisionAttribute = 5,
AttributesDomain = new double[][]
{
null, // no need for "id"
new double[]{ 0, 1, 2 },
new double[]{ 0, 1, 2 },
new double[]{ 0, 1, 2 },
new double[]{ 0, 1, 2 },
new double[]{ 0, 1, 2 },
},
};
Console.WriteLine("Running...");
QuickReductProcessor processor = new QuickReductProcessor(infSys);
// Show results
Console.WriteLine("\nAll reducts: ");
for(int i = 0; i < processor.Results.Count; i++ ) {
var rs = processor.Results[i];
Console.WriteLine((i+1).ToString() + ". {" + string.Join(",", rs.OrderBy(x => x) ) + "}");
}
Console.WriteLine("\nMinimal reducts: ");
var miRd = processor.MinimalReducts.ToArray();
for (int i = 0; i < miRd.Length; i++)
{
var rs = miRd[i];
Console.WriteLine((i + 1).ToString() + ". {" + string.Join(",", rs.OrderBy(x => x)) + "}");
}
Console.WriteLine("\nCore attributes: " + string.Join(",", processor.CoreAttrs));
// Keep the console open in debug mode.
Console.WriteLine(System.Environment.NewLine);
Console.WriteLine("Press any key to exit");
Console.ReadKey();
}
public Chi2Discretizer(DecisionSystem _DS)
{
this.DS = _DS;
this.DecisionClasses = DS.AttributesDomain[DS.DecisionAttribute].OrderBy(x => x).ToList();
this.DegreeOfFreedom = DecisionClasses.Count - 1;
this.InitCache();
this.Delta = this.CalcCurrentInconsistencyRate();
//TODO: If Ei < 5 , raise an exception
}
public static void RunTest()
{
// Prepare test data
var infSys = new DecisionSystem
{
// KhuPN, Hong - "Datamining based on Rough set theory" pg. 8
Universe = new double?[][]
{
// id 1 2 3 4 5
new double?[] {1, 2, 1, 0, 1, 0},
new double?[] {2, 2, 0, 1, 0, 1},
new double?[] {3, 1, 0, 0, 1, 0},
new double?[] {4, 0, 1, 0, 0, 1},
new double?[] {5, 0, 0, 0, 1, 1},
new double?[] {6, 2, 0, 0, 1, 0},
new double?[] {7, 0, 0, 1, 0, 1},
new double?[] {8, 2, 1, 1, 0, 1}
},
ConditionAttributes = new[] {1,2, 3, 4},
DecisionAttributes = new[] {5},
DecisionAttribute = 5,
AttributesDomain = new double[][]
{
null, // no need for "id"
new double[] {0, 1, 2},
new double[] {0, 1},
new double[] {0, 1},
new double[] {0, 1},
new double[] {0, 1},
},
};
var r=new Rules(infSys,new int[]{1,2,3,4});
//Output
foreach (var discernibilityMatrix in r.DiscernMatrix)
{
Console.WriteLine("Attribute: "+discernibilityMatrix.RuleValue);
Console.WriteLine("Prime Impicants:");
foreach (var primeImplicant in discernibilityMatrix.PrimeImplicants)
{
foreach (var pairId in primeImplicant)
{
Console.Write(pairId.AID+":"+pairId.VID+" ; ");
}
Console.WriteLine();
}
Console.WriteLine();
}
Console.WriteLine("Press any key to exit");
Console.ReadKey();
}
public DecisionSystem GetDecisionSystem(List<AttributeDefinition> attributeDefinitions, string tableName)
{
//var infSys = new DecisionSystem
//{ // KhuPN - Table 2 page 5
// Universe = new double?[][]
// { // id 1 2 3 4
// new double?[]{ 1, 0, 1, 1, 1 },
// new double?[]{ 2, 1, 0, 1, 1 },
// new double?[]{ 3, 1, 1, 2, 1 },
// new double?[]{ 4, 0, 1, 0, 0 },
// new double?[]{ 5, 1, 0, 1, 0 },
// new double?[]{ 6, 0, 1, 2, 1 },
// },
// ConditionAttributes = new[] { 1, 2, 3 },
// DecisionAttributes = new[] { 4 },
// DecisionAttribute = 4,
// AttributesDomain = new double[][]
// {
// null, // no need for "id"
// new double[]{ 0, 1 },
// new double[]{ 0, 1 },
// new double[]{ 0, 1, 2 },
// new double[]{ 0, 1 },
// },
//};
int[] conditionAttributes;
int[] decisionAttributes;
double?[][] universe;
double[][] attributeDomain;
int idIndex = 0;
DataRepository.GetDataForDecisionSystem(attributeDefinitions, tableName, out universe, out attributeDomain, out decisionAttributes, out conditionAttributes, ref idIndex);
DecisionSystem ds = new DecisionSystem
{
Universe = universe,
ConditionAttributes = conditionAttributes,
DecisionAttributes = decisionAttributes,
DecisionAttribute = decisionAttributes[0],
AttributesDomain = attributeDomain
};
return ds;
}
public static void perfom()
{
//var chi2val = SpecialFunction.chisq(1, 0.5);
//return;
// load file
var infSys = new DecisionSystem();
string filename = @"data/iris.csv";
//string filename = @"data/input1.csv";
StreamReader reader = new StreamReader(filename);
infSys.ConditionAttributes = new int[] { 1, 2, 3, 4 };
infSys.DecisionAttribute = 5;
// Universe
reader.ReadLine(); // ignore header line
List<double?[]> objs = new List<double?[]>();
List<double?> obj;
int idTracker = 0;
Dictionary<int, List<double>> attrDomain = new Dictionary<int,List<double>>();
for(int i = 0; i < 6; i ++)
attrDomain.Add(i, new List<double>());
while (!reader.EndOfStream)
{
string line = reader.ReadLine();
obj = new List<double?>();
obj.Add(++idTracker);
var numbers = line.Split(',')
.Select((x, i) => new {str = x, idx = i})
.Where(o => o.idx < 5)
.Select(t => (double?)Double.Parse(t.str))
.ToList();
obj.AddRange(numbers);
objs.Add(obj.ToArray());
// get value domain of all attrs
// ignore the id (i = 0) of object
for (int i = 1; i < 6; i++) {
if(obj[i] != null && !attrDomain[i].Contains((double)obj[i]) )
attrDomain[i].Add((double)obj[i]);
}
}
infSys.Universe = objs.ToArray();
infSys.AttributesDomain = attrDomain
.OrderBy(p => p.Key)
.Select(p => p.Value.ToArray())
.ToArray();
reader.Close();
printDS(infSys, @"data/inputDS.txt");
// get begin inconsistency rate
var apprSpace = new StandardApproximationSpace(infSys, infSys.ConditionAttributes);
double rate = apprSpace.InConsistencyRate();
// Process data
Chi2Discretizer dis = new Chi2Discretizer(infSys);
dis.Process();
// Get resulted DS
var processedDS = dis.BuildProcessedDS();
printDS(processedDS, @"data/processedDS.txt");
// get rate after processed
var apprSpace2 = new StandardApproximationSpace(processedDS, processedDS.ConditionAttributes);
double rate2 = apprSpace2.InConsistencyRate();
Console.WriteLine(rate2);
// Print report
StreamWriter writer = new StreamWriter(@"data/report.txt", false);
writer.WriteLine("Data file: " + filename);
writer.WriteLine("- Item Count: " + infSys.Universe.Length);
writer.WriteLine("- Original condition attributes: " + string.Join(",", infSys.ConditionAttributes));
writer.WriteLine("- Selected attributes (reduct): " + string.Join(",", dis.SelectedConditionAttrs));
writer.WriteLine("- Original domain count: ");
foreach (int attr in infSys.ConditionAttributes) {
writer.Write(attr + ": ");
writer.WriteLine(infSys.AttributesDomain[attr].Length.ToString());
}
writer.WriteLine("- Processed domain count: ");
foreach (int attr in processedDS.ConditionAttributes)
{
writer.Write(attr + ": ");
writer.WriteLine(processedDS.AttributesDomain[attr].Length.ToString());
}
writer.WriteLine("- Removed attributes: " + string.Join(",", dis.RemovedConditionAttrs));
writer.WriteLine("- Original in-consistency rate: " + rate);
writer.WriteLine("- Processed in-consistency rate: " + rate2);
if (rate2 > 0)
{
var inConClasses = apprSpace2.IndiscernibilityClasses()
.Where(X => ApproximationSpace.GetAllDecisionClasses(X, processedDS.DecisionAttribute).Count() > 1);
writer.WriteLine("- All in-consistent classes:");
int count = 0;
foreach(var klass in inConClasses){
count++;
writer.WriteLine("\t+ Class: " + count);
foreach (var o in klass) {
writer.WriteLine("\t\t" + string.Join(",", o));
}
}
}
writer.Close();
}
public static void printDS(DecisionSystem ds, string fileName)
{
StreamWriter writer = new StreamWriter(fileName);
foreach (var obj in ds.Universe) {
var line = obj
.Select(x => x.ToString())
.Aggregate((joinstr, str) => joinstr + "\t" + str);
writer.WriteLine(line);
}
writer.Close();
}
// check whether new inconsistency rate will exceed the original rate
private bool WillNextMergedIntervalIncreaseInConsisRate(int targetAttr, List<double> chi2Values)
{
int firstIntervalIdx = chi2Values.IndexOf(chi2Values.Min());
double replacedInterval = this.processedAttrDomains[targetAttr][firstIntervalIdx];
double removedInterval = this.processedAttrDomains[targetAttr][firstIntervalIdx + 1];
var newUniverse = this.crnUniverse.Select(x => (double?[])x.Clone()).ToArray();
for (int i = 0; i < newUniverse.Count(); i++)
{
var newObj = newUniverse[i];
//FIXME: for test
if (newObj[0] == 77)
Console.WriteLine();
if (newObj[targetAttr] == removedInterval)
newObj[targetAttr] = replacedInterval;
}
var newDS = new DecisionSystem(newUniverse.ToArray(), this.SelectedConditionAttrs, this.DS.DecisionAttribute);
var newApprSpace = new StandardApproximationSpace(newDS, this.SelectedConditionAttrs);
var newRate = newApprSpace.InConsistencyRate();
return (newRate > this.Delta);
}
public DecisionSystem BuildProcessedDS()
{
DecisionSystem newDS = new DecisionSystem();
newDS.AttributesDomain = this.processedAttrDomains
.OrderBy(pair => pair.Key)
.Select(pair => pair.Value.ToArray())
.ToArray();
newDS.ConditionAttributes = this.SelectedConditionAttrs;
newDS.DecisionAttribute = this.DS.DecisionAttribute;
newDS.DecisionAttributes = this.DS.DecisionAttributes;
newDS.Universe = this.crnUniverse;
return newDS;
}
public StandardApproximationSpace(DecisionSystem infoSys, IEnumerable<int> B)
: base(infoSys, IndiscernibilityRelation.Equivalence(B), RoughInclusion.Standard)
{
if (!B.Any()) throw new ArgumentException("Should have at least 1 condition attributes", "B");
Attrs = B;
}