/// <summary>
/// Calculates the maximum distance for all values for the provided attribute
/// using the provided ISimilarityMeasure
/// </summary>
/// <param name="attribute">The attribute to retrieve the maximum distance for</param>
/// <param name="measure">The ISimilarityMeasure to use to calculate the distance</param>
/// <returns>the maximum distance value for all values for the given attribute</returns>
private double CalculateMaxDistance(Data.Attributes.Attribute attribute, ISimilarityMeasure measure)
{
// Ensure that the global collection contains the attribute
if (!GlobalAttributeCollection.GetInstance(this.scope).ContainsAttribute(attribute))
{
return(0);
}
double maxDistance = double.MinValue;
List <string> values = new List <string>(GlobalAttributeCollection.GetInstance(this.scope).GetAttributeValues(attribute));
//TODO: IS THERE A MORE EFFICIENT METHOD FOR HANDLING THIS??
for (int i = 0; i < values.Count; i++)
{
// get the value for this item
string a = values[i];
// Compare the current item to all other items in the list
for (int j = 0; j < i; j++)
{
// get the value for this item
string b = values[j];
double?distance = measure.CalculateDistance(a, b);
if (distance != null)
{
maxDistance = Math.Max(maxDistance, (double)distance);
}
}
}
return(maxDistance);
}
/// <summary>
/// Calculates the maximum distance for all values for the provided attribute
/// using the provided ISimilarityMeasure
/// </summary>
/// <param name="attribute">The attribute to retrieve the maximum distance for</param>
/// <param name="measure">The ISimilarityMeasure to use to calculate the distance</param>
/// <returns>the maximum distance value for all values for the given attribute</returns>
private double CalculateMaxDistance(Data.Attributes.Attribute attribute, ISimilarityMeasure measure)
{
// Ensure that the global collection contains the attribute
if (!GlobalAttributeCollection.GetInstance(this.scope).ContainsAttribute(attribute))
return 0;
double maxDistance = double.MinValue;
List<string> values = new List<string>(GlobalAttributeCollection.GetInstance(this.scope).GetAttributeValues(attribute));
//TODO: IS THERE A MORE EFFICIENT METHOD FOR HANDLING THIS??
for (int i = 0; i < values.Count; i++)
{
// get the value for this item
string a = values[i];
// Compare the current item to all other items in the list
for (int j = 0; j < i; j++)
{
// get the value for this item
string b = values[j];
double? distance = measure.CalculateDistance(a, b);
if (distance != null)
maxDistance = Math.Max(maxDistance, (double)distance);
}
}
return maxDistance;
}
/// <summary>
/// Returns a list of tuples that contain the calculated distances
/// and the frequency of those distances
/// </summary>
/// <param name="attributeName">The name of the attribute that
/// distances are being calculated for</param>
/// <param name="measure">The similarity measure to be used</param>
/// <returns>a collection of distances and the number of times
/// that those distances occur</returns>
private List<Tuple<double, int>> CalculateDistances(string attributeName, ISimilarityMeasure measure)
{
if (!GlobalAttributeCollection.GetInstance(this.scope).ContainsAttribute(attributeName))
return null;
List<Tuple<double, int>> distances = new List<Tuple<double, int>>();
double frequencyTotal = 0;
int nodeCount = 0;
// Get the values for the attribute
List<string> attributeValues = new List<string>(GlobalAttributeCollection.GetInstance(this.scope).GetAttributeValues(attributeName));
// Loop over all the attribute values
for (int i = 0; i <= attributeValues.Count - 1; i++)
{
// Get the frequency at which the source attribute value occurrs
int sourceFrequency = GlobalAttributeCollection.GetInstance(this.scope).GetFrequency(attributeName, attributeValues[i]);
nodeCount += sourceFrequency;
// Compare the current attribute value to all other
// attribute values
for (int j = 0; j < i; j++)
{
// Compute the distance for the two values
double? distance = measure.CalculateDistance(attributeValues[i], attributeValues[j]);
if (distance != null)
{
// Get the frequency at which the target attribute value occurrs
int targetFrequency = GlobalAttributeCollection.GetInstance(this.scope).GetFrequency(attributeName, attributeValues[j]);
distances.Add(Tuple.Create<double, int>(distance.Value, sourceFrequency * targetFrequency));
// Keep a running total of the frequencies
frequencyTotal += sourceFrequency * targetFrequency;
}
}
}
// Since we only loop over unique attribute values we never make
// the comparisons against nodes where the values would be the
// same. We need to determine if this case has occurred and insert
// the appropriate number of zero distance items.
// Use binomial function to determine the number of possible combinations
//double combinations = (MathUtils.LogFactorial(nodeCount) / (2 * (MathUtils.LogFactorial(nodeCount - 2))));
double combinations = Math.Exp(MathUtils.LogFactorial(nodeCount) - MathUtils.LogFactorial(2) - MathUtils.LogFactorial(nodeCount - 2));
// Add in all the zero distance items that we need
for (int i = 1; i <= combinations - frequencyTotal; i++)
{
distances.Add(Tuple.Create<double, int>(0, 1));
}
//foreach (Tuple<double, int> distanceCount in distances)
//{
// System.Diagnostics.Debug.WriteLine("[{0},{1}]", distanceCount.Item1, distanceCount.Item2);
//}
return distances;
}
/// <summary>
/// Returns a list of tuples that contain the calculated distances
/// and the frequency of those distances
/// </summary>
/// <param name="attributeName">The name of the attribute that
/// distances are being calculated for</param>
/// <param name="measure">The similarity measure to be used</param>
/// <returns>a collection of distances and the number of times
/// that those distances occur</returns>
private List <Tuple <double, int> > CalculateDistances(string attributeName, ISimilarityMeasure measure)
{
if (!GlobalAttributeCollection.GetInstance(this.scope).ContainsAttribute(attributeName))
{
return(null);
}
List <Tuple <double, int> > distances = new List <Tuple <double, int> >();
double frequencyTotal = 0;
int nodeCount = 0;
// Get the values for the attribute
List <string> attributeValues = new List <string>(GlobalAttributeCollection.GetInstance(this.scope).GetAttributeValues(attributeName));
// Loop over all the attribute values
for (int i = 0; i <= attributeValues.Count - 1; i++)
{
// Get the frequency at which the source attribute value occurrs
int sourceFrequency = GlobalAttributeCollection.GetInstance(this.scope).GetFrequency(attributeName, attributeValues[i]);
nodeCount += sourceFrequency;
// Compare the current attribute value to all other
// attribute values
for (int j = 0; j < i; j++)
{
// Compute the distance for the two values
double?distance = measure.CalculateDistance(attributeValues[i], attributeValues[j]);
if (distance != null)
{
// Get the frequency at which the target attribute value occurrs
int targetFrequency = GlobalAttributeCollection.GetInstance(this.scope).GetFrequency(attributeName, attributeValues[j]);
distances.Add(Tuple.Create <double, int>(distance.Value, sourceFrequency * targetFrequency));
// Keep a running total of the frequencies
frequencyTotal += sourceFrequency * targetFrequency;
}
}
}
// Since we only loop over unique attribute values we never make
// the comparisons against nodes where the values would be the
// same. We need to determine if this case has occurred and insert
// the appropriate number of zero distance items.
// Use binomial function to determine the number of possible combinations
//double combinations = (MathUtils.LogFactorial(nodeCount) / (2 * (MathUtils.LogFactorial(nodeCount - 2))));
double combinations = Math.Exp(MathUtils.LogFactorial(nodeCount) - MathUtils.LogFactorial(2) - MathUtils.LogFactorial(nodeCount - 2));
// Add in all the zero distance items that we need
for (int i = 1; i <= combinations - frequencyTotal; i++)
{
distances.Add(Tuple.Create <double, int>(0, 1));
}
//foreach (Tuple<double, int> distanceCount in distances)
//{
// System.Diagnostics.Debug.WriteLine("[{0},{1}]", distanceCount.Item1, distanceCount.Item2);
//}
return(distances);
}