public void CalculateNormativeFit(Context currentContext, Simulation simulation)
{
int numDimensions = currentContext.relevantCharacteristcs.Count;
for (int c = 0; c < clusterMeans.Count; c++)
{
var clusterMean = clusterMeans[c];
float bestDistance = float.MaxValue;
foreach (var socialGroup in knowledgeBase)
{
bool validGroup = true;
float distance = 0.0f;
for (int i = 0; i < numDimensions; i++)
{
float groupValue = 0.0f;
if (!socialGroup.characteristics.TryGetValue(currentContext.relevantCharacteristcs[i], out groupValue))
{
validGroup = false;
break;
}
float weight = currentContext.relevantCharacteristcs[i].weight;
float difference = (clusterMean.mean[i] - groupValue) * weight;
distance += difference * difference;
}
if (!validGroup)
{
continue;
}
distance = (float)Math.Sqrt((double)distance);
if (distance < bestDistance)
{
bestDistance = distance;
clusterMean.matchedGroup = socialGroup;
}
}
// ad-hoc group
if (bestDistance > simulation.normativeMatchDistance * currentContext.GetMaxDistance())
{
numAdhocGroups++;
SocialGroup adhocGroup = new SocialGroup();
adhocGroup.isAdHoc = true;
adhocGroup.characteristics = new Dictionary <Characteristic, float>();
adhocGroup.name = "Group " + numAdhocGroups;
for (int i = 0; i < numDimensions; i++)
{
adhocGroup.characteristics.Add(currentContext.relevantCharacteristcs[i], clusterMean.mean[i]);
}
adhocGroup.CalculateAccessibility(this);
knowledgeBase.Add(adhocGroup);
clusterMean.matchedGroup = adhocGroup;
}
clusterMeans[c] = clusterMean;
}
}
/*
* public void NormalizeAccessibility()
* {
* float totalAccessibility = 0.0f;
*
* foreach (var socialGroup in knowledgeBase)
* {
* totalAccessibility += socialGroup.accessibility;
* }
*
* if (totalAccessibility == 0.0f)
* {
* foreach (var socialGroup in knowledgeBase)
* {
* socialGroup.accessibility = 1.0f / knowledgeBase.Count;
* }
* }
* else
* {
* foreach (var socialGroup in knowledgeBase)
* {
* socialGroup.accessibility /= totalAccessibility;
* }
* }
* }*/
public void CalculateComparativeFit(Context currentContext, Simulation simulation)
{
salience = minimalSalienceThreshold;
foreach (var socialGroup in knowledgeBase)
{
float groupComparativeFit;
int numDimensions = currentContext.relevantCharacteristcs.Count;
int selfCluster = GetSelfCluster();
float meanDistance = 0.0f, meanDispersion = 0.0f;
bool validGroup = true;
for (int c = 0; c < clusterMeans.Count; c++)
{
if (c == selfCluster)
{
continue;
}
var clusterMean = clusterMeans[c];
float distance = 0.0f;
for (int i = 0; i < numDimensions; i++)
{
// Comparative distance with visible groups or matched groups
float groupValue = 0.0f, selfGroupValue = 0.0f;
clusterMean.matchedGroup.characteristics.TryGetValue(currentContext.relevantCharacteristcs[i], out groupValue);
if (!socialGroup.characteristics.TryGetValue(currentContext.relevantCharacteristcs[i], out selfGroupValue))
{
validGroup = false;
break;
}
float weight = currentContext.relevantCharacteristcs[i].weight;
float difference = (groupValue - selfGroupValue) * weight;
distance += difference * difference;
}
if (!validGroup)
{
break;
}
meanDistance += (float)Math.Sqrt((double)distance) / currentContext.GetMaxDistance();
meanDispersion += clusterMean.dispersion;
}
if (!validGroup)
{
continue;
}
if (clusterMeans.Count > 1)
{
meanDistance /= clusterMeans.Count - 1;
meanDispersion /= clusterMeans.Count - 1;
groupComparativeFit = simulation.comparativeFitAlfa * meanDistance + (1.0f - simulation.comparativeFitAlfa) *
(simulation.comparativeFitBeta * (1.0f - clusterMeans[selfCluster].dispersion) + (1.0f - simulation.comparativeFitBeta) * meanDispersion);
}
else
{
meanDistance = 0.0f;
groupComparativeFit = 0.0f;
}
float groupSalience = groupComparativeFit * socialGroup.accessibility;
socialGroup.salience = groupSalience;
if (groupSalience > salience)
{
selfGroup = socialGroup;
salience = groupSalience;
comparativeFit = groupComparativeFit;
groupMeanDistance = meanDistance;
}
}
}