public IEnumerator <ListPQNode <Cluster> > GetEnumerator()
{
ListPQNode <Cluster> current = head;
while (current != null)
{
yield return(current);
current = current.prev;
}
}
public void removeMin()
{
if (N == 1)
{
N--;
head = null;
tail = null;
return;
}
tail.next.prev = null;
tail = tail.next;
N--;
}
public void insertMax(Cluster c)
{
if (N == 0)
{
head = new ListPQNode <Cluster>(c);
tail = head;
N++;
return;
}
ListPQNode <Cluster> n = new ListPQNode <Cluster>(c);
head.next = n;
n.prev = head;
head = n;
N++;
if (N > M)
{
removeMin();
}
}
public void setMax(ListPQNode <Cluster> n)
{
if (N == 1 || head == n)
{
return;
}
else if (tail == n)
{
tail = n.next;
tail.prev = null;
}
else
{
ListPQNode <Cluster> left = n.prev;
ListPQNode <Cluster> right = n.next;
left.next = right;
right.prev = left;
}
//set selected node to max
head.next = n;
n.prev = head;
head = n;
}
/// <summary>
/// Checks whether a record belongs in a cluster using string comparison
/// </summary>
private bool compareRecordToCluster(Record queryRecord, Cluster cluster, double tolerance, ListPQNode <Cluster> node, bool scanDates, bool scanDescriptions, double namePrecision, double datePrecision, double descriptionPrecision)
{
bool result = false;
foreach (Record clusterRecord in cluster.getRecords())
{ //check if record is similar enough to record in cluster to be added
//TODO change logic so that it only calculates all three measures if search enhance is on
bool similarityFail = false;
double totalSimilarity = 0;
int divisor = 1;
//first perform mandatory name check
double nameSimilarity = strComp.jaroWinklerCompare(queryRecord, clusterRecord);
if (nameSimilarity < namePrecision)
{
similarityFail = true;
}
double dateSimilarity = compareDates(queryRecord, clusterRecord);
if (scanDates && (dateSimilarity < datePrecision))
{
similarityFail = true;
divisor++;
}
double descriptionSimilarity = compareDescriptions(queryRecord, clusterRecord);
if (scanDescriptions && (descriptionSimilarity < descriptionPrecision))
{
similarityFail = true;
divisor++;
}
//calculate total similairty
//TODO smarter weighting
totalSimilarity = (nameSimilarity + dateSimilarity + descriptionPrecision) / divisor;
//if all three similarity checks succeeded, it's a match
if (!similarityFail)
{ //if yes, update the cluster
addRecordToCluster(queryRecord, cluster);
updatePQ(node); //and update the priority queue
result = true;
break;
}
else if (totalSimilarity < 0.4) //if the similarity is way off, don't bother checking the rest of the cluster
{
break;
}
}
return(result);
}
/// <summary>
/// Updates the PQ with the given cluster
/// </summary>
private void updatePQ(ListPQNode <Cluster> node)
{
listPQ.setMax(node);
}
/// <summary>
/// Checks whether record is in cluster, using pre-defined tolerance
/// </summary>
private bool compareRecordToClusterAuto(Record queryRecord, Cluster cluster, double tolerance, ListPQNode <Cluster> node, bool scanDates, bool scanDescriptions, List <String> ignoreList)
{
bool result = false;
foreach (Record r in cluster.getRecords())
{ //check if record is similar enough to record in cluster to be added
double nameWeight = 0; double dateWeight = 0; double descriptionWeight = 0; //will be used to determine how much weight to give to each field
double nameSimilarity = 0; double dateSimilarity = 0; double descriptionSimilarity = 0;
nameSimilarity = normalize(MIN_NAME_SIM, 1, strComp.jaroWinklerCompare(queryRecord, r));
//account for cases where one or both of the records are missing a date - ignore the field in calculation
if (queryRecord.getDate().Equals(new DateTime(1900, 1, 1)) || r.getDate().Equals(new DateTime(1900, 1, 1)))
{
scanDates = false;
}
else
{
dateSimilarity = normalize(0, MAX_DAYS, (MAX_DAYS - compareDates(queryRecord, r)));
}
//do the same for descriptions
if ((scanDescriptions == false) || IgnoreDescriptions(queryRecord, r, ignoreList))
{
scanDescriptions = false;
}
else
{
descriptionSimilarity = normalize(MIN_DESCRIPTION_SIM, MAX_DESCRIPTION_SIM, compareDescriptions(queryRecord, r));
}
calculateWeights(ref nameWeight, ref dateWeight, ref descriptionWeight, scanDates, scanDescriptions);
double similarity = (nameSimilarity * nameWeight) + (dateSimilarity * dateWeight) + (descriptionSimilarity * descriptionWeight);
if (similarity >= tolerance)
{ //if yes, update the cluster
addRecordToCluster(queryRecord, cluster);
updatePQ(node); //and update the priority queue
result = true;
break;
}
else if (similarity < tolerance / TOLERANCE_DISCARD_FACTOR) //if the similarity is way off, don't bother checking the rest of the cluster
{
break;
}
}
return(result);
}
