public List <ItemSet <T> > MinePatternsContaining(T item, GetMinSupportHandle getMinItemSetSupport)
{
List <ItemSet <T> > selection = new List <ItemSet <T> >();
MinePatternsContaining(mRoot, selection, item, getMinItemSetSupport);
return(selection);
}
public ItemSets <T> MinePatterns(IEnumerable <Transaction <T> > database, IList <T> domain, GetMinSupportHandle getMinItemSetSupport)
{
Dictionary <T, int> counts = new Dictionary <T, int>();
for (int i = 0; i < domain.Count; ++i)
{
counts[domain[i]] = 0;
}
foreach (Transaction <T> transaction in database)
{
for (int i = 0; i < domain.Count; ++i)
{
if (transaction.ContainsItem(domain[i]))
{
counts[domain[i]]++;
}
}
}
List <T> freqItems = new List <T>();
for (int i = 0; i < domain.Count; ++i)
{
var xxx = getMinItemSetSupport(new ItemSet <T>()
{
domain[i]
});
if (counts[domain[i]] >= getMinItemSetSupport(new ItemSet <T>()
{
domain[i]
}))
{
freqItems.Add(domain[i]);
}
}
T[] FList = freqItems.ToArray();
Array.Sort(FList, (i1, i2) =>
{
int comp = counts[i1].CompareTo(counts[i2]);
return(-comp);
});
FPTree <T> fpTree = new FPTree <T>();
int dbSize = 0;
foreach (Transaction <T> transaction in database)
{
dbSize++;
List <T> orderedFreqItems = new List <T>();
for (int i = 0; i < FList.Length; ++i)
{
if (transaction.ContainsItem(FList[i]))
{
orderedFreqItems.Add(FList[i]);
}
}
fpTree.AddOrderedFreqItems(orderedFreqItems);
}
fpTree.DbSize = dbSize;
ItemSets <T> allItemSets = new ItemSets <T>();
for (int i = FList.Length - 1; i >= 0; i--)
{
T item = FList[i];
List <ItemSet <T> > fis = fpTree.MinePatternsContaining(item, getMinItemSetSupport(new ItemSet <T>()
{
item
}));
fpTree.RemoveFromLeaves(item);
allItemSets.AddRange(fis);
}
return(allItemSets);
}
protected void MinePatternsContaining(FPTreeNode <T> node, List <ItemSet <T> > selection, T targetItem, GetMinSupportHandle getMinItemSetSupport)
{
if (node.IsLeaf)
{
if (node.Item.Equals(targetItem))
{
ItemSet <T> fis = new ItemSet <T>();
fis = node.GetPath();
fis.TransactionCount = node.Count;
fis.DbSize = mDbSize;
if (node.Count >= getMinItemSetSupport(fis) * mDbSize)
{
selection.Add(fis);
}
}
return;
}
for (int i = 0; i < node.ChildCount; ++i)
{
MinePatternsContaining(node.GetChild(i), selection, targetItem, getMinItemSetSupport);
}
}
protected ItemSets <T> GenerateLargeItemSets(List <Transaction <T> > partition, GetMinSupportHandle getMinItemSetSupport, IList <T> domain)
{
ItemSets <T> Fk = new ItemSets <T>();
for (int i = 0; i < domain.Count; ++i)
{
T item = domain[i];
ItemSet <T> itemset = new ItemSet <T>()
{
item
};
for (int j = 0; j < partition.Count; ++j)
{
if (partition[j].ContainsItemSet(itemset))
{
long tid = partition[j].ID;
itemset.TransactionIDList.Add(tid);
}
}
if (itemset.TransactionIDList.Count >= getMinItemSetSupport(itemset) * partition.Count)
{
Fk.Add(itemset);
}
}
int k = 1;
ItemSets <T> allFrequentItemSets = new ItemSets <T>();
allFrequentItemSets.AddRange(Fk);
while (Fk.Count > 0)
{
ItemSets <T> Fkp1 = new ItemSets <T>();
//do self-join
for (int i = 0; i < Fk.Count; ++i)
{
for (int j = 0; j < Fk.Count; ++j)
{
if (i == j)
{
continue;
}
bool canJoin = true;
for (int l = 0; l < k - 1; ++l)
{
if (Fk[i][l].CompareTo(Fk[j][l]) != 0)
{
canJoin = false;
break;
}
}
if (canJoin)
{
if (CanJoin(Fk[i], Fk[j][k - 1]))
{
ItemSet <T> c = Union(Fk[i], Fk[j]);
if (c.TransactionIDList.Count >= getMinItemSetSupport(c) * partition.Count)
{
Fkp1.Add(c);
}
}
}
}
}
allFrequentItemSets.AddRange(Fkp1);
Fk = Fkp1;
k++;
}
return(allFrequentItemSets);
}
public virtual ItemSets <T> MinePatterns(IEnumerable <Transaction <T> > database, GetMinSupportHandle getMinItemSetSupport, IList <T> domain, int partitionCount)
{
HashSet <ItemSet <T> > candidates = new HashSet <ItemSet <T> >();
for (int i = 0; i < partitionCount; ++i)
{
List <Transaction <T> > partition = ReadInPartition(i, database);
ItemSets <T> fis = GenerateLargeItemSets(partition, getMinItemSetSupport, domain);
foreach (ItemSet <T> itemset in fis)
{
candidates.Add(itemset);
}
}
int dbSize = 0;
for (int i = 0; i < partitionCount; ++i)
{
List <Transaction <T> > partition = ReadInPartition(i, database);
dbSize += partition.Count;
foreach (ItemSet <T> itemset in candidates)
{
itemset.TransactionCount += GetCount(partition, itemset);
}
}
foreach (ItemSet <T> itemset in candidates)
{
itemset.DbSize = dbSize;
}
ItemSets <T> C = new ItemSets <T>();
foreach (ItemSet <T> itemset in candidates)
{
if (itemset.Support >= getMinItemSetSupport(itemset))
{
C.Add(itemset);
}
}
return(C);
}
public virtual ItemSets <T> MinePatterns(GetCountHandle updateItemSetSupport, GetMinSupportHandle getMinItemSetSupport, IList <T> domain)
{
ItemSets <T> Fk = new ItemSets <T>();
List <ItemSet <T> > itemsetSup = new List <ItemSet <T> >();
for (int i = 0; i < domain.Count; ++i)
{
T item = domain[i];
ItemSet <T> itemset = new ItemSet <T>()
{
item
};
itemset.TransactionCount = 0;
itemsetSup.Add(itemset);
}
updateItemSetSupport(itemsetSup);
foreach (ItemSet <T> itemset in itemsetSup)
{
if (itemset.Support >= getMinItemSetSupport(itemset))
{
Fk.Add(itemset);
}
}
int k = 1;
ItemSets <T> allFrequentItemSets = new ItemSets <T>();
allFrequentItemSets.AddRange(Fk);
while (Fk.Count > 0)
{
ItemSets <T> Fkp1 = new ItemSets <T>();
//do self-join
for (int i = 0; i < Fk.Count; ++i)
{
for (int j = 0; j < Fk.Count; ++j)
{
if (i == j)
{
continue;
}
bool canJoin = true;
for (int l = 0; l < k - 1; ++l)
{
if (Fk[i][l].CompareTo(Fk[j][l]) != 0)
{
canJoin = false;
break;
}
}
if (canJoin)
{
if (CanJoin(Fk[i], Fk[j][k - 1]))
{
ItemSet <T> c = Fk[i].Clone();
c.Add(Fk[j][k - 1]);
Fkp1.Add(c);
}
}
}
}
updateItemSetSupport(Fkp1);
List <ItemSet <T> > fis = new List <ItemSet <T> >();
foreach (ItemSet <T> itemset in Fkp1)
{
if (itemset.Support >= getMinItemSetSupport(itemset))
{
fis.Add(itemset);
}
}
allFrequentItemSets.AddRange(fis);
Fk.Clear();
Fk.AddRange(fis);
k++;
}
return(allFrequentItemSets);
}
public virtual ItemSets <T> MinePatterns(IEnumerable <Transaction <T> > database, GetMinSupportHandle getMinItemSetSupport, IList <T> domain)
{
return(MinePatterns((itemsets) =>
{
foreach (ItemSet <T> itemset in itemsets)
{
itemset.TransactionCount = 0;
}
int dbSize = 0;
foreach (Transaction <T> transaction in database)
{
foreach (ItemSet <T> itemset in itemsets)
{
if (transaction.ContainsItemSet(itemset))
{
itemset.TransactionCount++;
}
}
dbSize++;
}
foreach (ItemSet <T> itemset in itemsets)
{
itemset.DbSize = dbSize;
}
}, getMinItemSetSupport, domain));
}