public k_foldCrossValidation(DataSetCross dataSet)
{
ruleSetList = new List <RuleSet>();
qualityList = new List <double>();
this.dataSet = dataSet;
//Parallel.For(0, dataSet.numberOfSets, i =>
//{
// List<Dictionary<string, string>> trainingSet = dataSet.GetTrainingSet(i);
// GreedyAlgorithm greedy = new GreedyAlgorithm(trainingSet, dataSet.headerRow, dataSet.decisionHeader);
// RuleSet temp = greedy.ruleSet;
// ruleSetList.Add(temp);
// qualityList.Add(GetQualityOfRuleSet(i, temp));
//}
//);
for (int i = 0; i < dataSet.numberOfSets; i++)
{
List <Dictionary <string, string> > trainingSet = dataSet.GetTrainingSet(i);
GreedyAlgorithm greedy = new GreedyAlgorithm(trainingSet, dataSet.headerRow, dataSet.decisionHeader);
RuleSet temp = greedy.ruleSet;
ruleSetList.Add(temp);
qualityList.Add(GetQualityOfRuleSet(i, temp));
}
SetAVGQuality();
SetBestQuality();
}
private Rule GetSubruleOfInacurrany(Rule rule, double inaccurancy, int inaccuracyOfTrainingTable)
{
List <Rule> subrulesList = new List <Rule>();
do
{
rule.decisionValue = GetMostCommonDecision(GetSubsetOfRule(rule, dataSet.trainingSet));
subrulesList.Add(rule.Copy());
rule.DeleteFirstCondition();
} while (rule.conditions.Count > 0);
List <double> inaccurancyList = new List <double>();
foreach (Rule subrule in subrulesList)
{
List <Dictionary <string, string> > subsetOfSubRule = GetSubsetOfRule(subrule.Copy(), dataSet.trainingSet);
inaccurancyList.Add(GreedyAlgorithm.getUncertaintyOfSubset(subsetOfSubRule, dataSet.decisionHeader) / inaccuracyOfTrainingTable);
}
int k = 0;
for (int i = 1; i < inaccurancyList.Count; i++)
{
if (inaccurancyList[i] != inaccurancyList[k] && inaccurancyList[i] <= inaccurancy)
{
k = i;
}
}
return(subrulesList[k]);
}
private void SplitSet(Random random)
{
foreach (var record in entireSet)
{
double rand = random.NextDouble();
if (rand < 0.3)
{
trainingSet.Add(record);
}
else if (rand < 0.5)
{
validationSet.Add(record);
}
else
{
testSet.Add(record);
}
}
if (GreedyAlgorithm.getUncertaintyOfSubset(trainingSet, decisionHeader) == 0 || GreedyAlgorithm.getUncertaintyOfSubset(validationSet, decisionHeader) == 0 || GreedyAlgorithm.getUncertaintyOfSubset(testSet, decisionHeader) == 0)
{
trainingSet = new List <Dictionary <string, string> >();
validationSet = new List <Dictionary <string, string> >();
testSet = new List <Dictionary <string, string> >();
SplitSet(random);
}
}
private RuleSet Pruning()
{
int inaccuracyOfTrainingTable = GreedyAlgorithm.getUncertaintyOfSubset(dataSet.trainingSet, dataSet.decisionHeader);
List <double> inaccuracyList = GetInaccuracyList(trainRuleSet.GetRuleSet(), inaccuracyOfTrainingTable);
List <RuleSet> ruleSetsList = new List <RuleSet>();
foreach (var i in inaccuracyList)
{
ruleSetsList.Add(new RuleSet(dataSet.headerRow, dataSet.decisionHeader));
}
for (int k = 0; k < ruleSetsList.Count; k++)
{
foreach (Rule rule in trainRuleSet.GetRuleSet())
{
ruleSetsList[k].AddRule(GetSubruleOfInacurrany(rule.Copy(), inaccuracyList[k], inaccuracyOfTrainingTable));
}
}
return(GetBestRuleSet(ruleSetsList));
}
private List <double> GetInaccuracyList(List <Rule> rules, int inaccuracyOfTrainingTable)
{
List <double> inaccuracyList = new List <double>();
foreach (Rule rule in rules)
{
Rule subrule = rule.Copy();
do
{
List <Dictionary <string, string> > subsetOfSubRule = GetSubsetOfRule(subrule, dataSet.trainingSet);
double inaccuracy = GreedyAlgorithm.getUncertaintyOfSubset(subsetOfSubRule, dataSet.decisionHeader) / inaccuracyOfTrainingTable;
if (!inaccuracyList.Contains(inaccuracy))
{
inaccuracyList.Add(inaccuracy);
}
subrule.DeleteFirstCondition();
} while (subrule.conditions.Count > 0);
}
inaccuracyList.Sort();
return(inaccuracyList);
}
private void SplitSet(Random random, double percent)
{
double threshold = percent / 100;
foreach (var record in entireSet)
{
double rand = random.NextDouble();
if (rand < threshold)
{
trainingSet.Add(record);
}
else
{
testSet.Add(record);
}
}
if (GreedyAlgorithm.getUncertaintyOfSubset(trainingSet, decisionHeader) == 0 || GreedyAlgorithm.getUncertaintyOfSubset(testSet, decisionHeader) == 0)
{
trainingSet = new List <Dictionary <string, string> >();
testSet = new List <Dictionary <string, string> >();
SplitSet(random, percent);
}
}
protected void Train()
{
GreedyAlgorithm greedyAlgorithm = new GreedyAlgorithm(dataSet.trainingSet, dataSet.headerRow, dataSet.decisionHeader);
trainRuleSet = greedyAlgorithm.ruleSet;
}
