void FrequentPatternGroup::frequentPatternGroup(std::vector<AdjacenceListsGRAPH> & queryGraphVector, std::vector<std::vector<int>> & similiarQueryGroups)
{
std::map<string, vector<int>> TLSInvertedIndex;
std::vector<Transaction> transactions;
for (std::vector<AdjacenceListsGRAPH>::iterator queryIterator = queryGraphVector.begin(); queryIterator != queryGraphVector.end(); queryIterator++) {
// each query is treated as a transaction
transactions.push_back(std::vector<string>());
Transaction & newTransaction = transactions[transactions.size() - 1];
for (std::map<TLSequence, std::vector<std::vector<int>>>::iterator tlsSeqIterator = queryIterator->getTLSequenceMap()->begin(); tlsSeqIterator != queryIterator->getTLSequenceMap()->end(); tlsSeqIterator++) {
string tlsQuanti = std::to_string(tlsSeqIterator->first.start);
tlsQuanti += "_";
tlsQuanti += std::to_string(tlsSeqIterator->first.pivot);
tlsQuanti += "_";
tlsQuanti += std::to_string(tlsSeqIterator->first.end);
tlsQuanti += "_";
for (int i = 0; i < tlsSeqIterator->second.size(); i++) {
string tlsIndexKey = tlsQuanti + std::to_string(i);
std::map<string, vector<int>>::iterator invertexIndexIter = TLSInvertedIndex.find(tlsIndexKey);
if (invertexIndexIter == TLSInvertedIndex.end()) {
vector<int> tlsIndexValue;
tlsIndexValue.push_back(queryIterator->graphId);
TLSInvertedIndex.insert(std::pair<string, vector<int>>(tlsIndexKey, tlsIndexValue));
}
else {
invertexIndexIter->second.push_back(queryIterator->graphId);
}
// we use a format i.e. "start_pivot_end_1" to distinuish each tls, thus we can can consider the number of the same tls of each query graph
newTransaction.push_back(tlsIndexKey);
}
}
}
const unsigned minimum_support_treshold = 2;
const FPTree fptree{ transactions, minimum_support_treshold };
std::set<Pattern> patterns = fptree_growth(fptree);
cout << "Frequent TLS Size: " << patterns.size() << endl;
for (std::set<Pattern>::iterator patternIterator = patterns.begin(); patternIterator != patterns.end(); patternIterator++) {
cout << patternIterator->second << " ";
for (std::set<Item>::iterator itemIterator = patternIterator->first.begin(); itemIterator != patternIterator->first.end(); itemIterator++) {
cout<<(*itemIterator) << " ";
}
cout << endl;
}
}
std::set<Pattern> fptree_growth(const FPTree& fptree) {
if ( fptree.empty() ) { return std::set<Pattern>{}; }
if ( contains_single_path( fptree ) ) {
// generate all possible combinations of the items in the tree
std::set<Pattern> single_path_patterns;
// for each node in the tree
assert( fptree.root->children.size() == 1 );
std::shared_ptr<FPNode> curr_fpnode = fptree.root->children.front();
while ( curr_fpnode ) {
const Item& curr_fpnode_item = curr_fpnode->item;
const unsigned curr_fpnode_frequency = curr_fpnode->frequency;
// add a pattern formed only by the item of the current node
Pattern new_pattern = { { curr_fpnode_item }, curr_fpnode_frequency };
single_path_patterns.insert( new_pattern );
// create a new pattern by adding the item of the current node to each pattern generated until now
for ( const Pattern& pattern : single_path_patterns ) {
Pattern new_pattern{ pattern };
new_pattern.first.insert( curr_fpnode_item );
assert( curr_fpnode_frequency <= pattern.second );
new_pattern.second = curr_fpnode_frequency;
single_path_patterns.insert( new_pattern );
}
// advance to the next node until the end of the tree
assert( curr_fpnode->children.size() <= 1 );
if ( curr_fpnode->children.size() == 1 ) { curr_fpnode = curr_fpnode->children.front(); }
else { curr_fpnode = nullptr; }
}
return single_path_patterns;
}
else {
// generate conditional fptrees for each different item in the fptree, then join the results
std::set<Pattern> multi_path_patterns;
// for each item in the fptree
for ( const auto& pair : fptree.header_table ) {
const Item& curr_item = pair.first;
// build the conditional fptree relative to the current item
// start by generating the conditional pattern base
std::vector<TransformedPrefixPath> conditional_pattern_base;
// for each path in the header_table (relative to the current item)
std::shared_ptr<FPNode> path_starting_fpnode = pair.second;
while ( path_starting_fpnode ) {
// construct the transformed prefix path
// each item in th transformed prefix path has the same frequency (the frequency of path_starting_fpnode)
const unsigned path_starting_fpnode_frequency = path_starting_fpnode->frequency;
std::shared_ptr<FPNode> curr_path_fpnode = path_starting_fpnode->parent;
// check if curr_path_fpnode is already the root of the fptree
if ( curr_path_fpnode->parent ) {
// the path has at least one node (excluding the starting node and the root)
TransformedPrefixPath transformed_prefix_path{ {}, path_starting_fpnode_frequency };
while ( curr_path_fpnode->parent ) {
assert( curr_path_fpnode->frequency >= path_starting_fpnode_frequency );
transformed_prefix_path.first.push_back( curr_path_fpnode->item );
// advance to the next node in the path
curr_path_fpnode = curr_path_fpnode->parent;
}
conditional_pattern_base.push_back( transformed_prefix_path );
}
// advance to the next path
path_starting_fpnode = path_starting_fpnode->node_link;
}
// generate the transactions that represent the conditional pattern base
std::vector<Transaction> conditional_fptree_transactions;
for ( const TransformedPrefixPath& transformed_prefix_path : conditional_pattern_base ) {
const std::vector<Item>& transformed_prefix_path_items = transformed_prefix_path.first;
const unsigned transformed_prefix_path_items_frequency = transformed_prefix_path.second;
Transaction transaction;
for ( const Item& item : transformed_prefix_path_items ) { transaction.push_back( item ); }
// add the same transaction transformed_prefix_path_items_frequency times
for ( int i = 0; i < transformed_prefix_path_items_frequency; ++i ) { conditional_fptree_transactions.push_back( transaction ); }
}
// build the conditional fptree relative to the current item with the transactions just generated
const FPTree conditional_fptree( conditional_fptree_transactions, fptree.minimum_support_treshold );
// call recursively fptree_growth on the conditional fptree (empty fptree: no patterns)
std::set<Pattern> conditional_patterns = fptree_growth( conditional_fptree );
// construct patterns relative to the current item using both the current item and the conditional patterns
std::set<Pattern> curr_item_patterns;
// the first pattern is made only by the current item
// compute the frequency of this pattern by summing the frequency of the nodes which have the same item (follow the node links)
unsigned curr_item_frequency = 0;
std::shared_ptr<FPNode> fpnode = pair.second;
while ( fpnode ) {
curr_item_frequency += fpnode->frequency;
fpnode = fpnode->node_link;
}
// add the pattern as a result
Pattern pattern{ { curr_item }, curr_item_frequency };
curr_item_patterns.insert( pattern );
// the next patterns are generated by adding the current item to each conditional pattern
for ( const Pattern& pattern : conditional_patterns ) {
Pattern new_pattern{ pattern };
new_pattern.first.insert( curr_item );
assert( curr_item_frequency >= pattern.second );
new_pattern.second = pattern.second;
curr_item_patterns.insert( { new_pattern } );
}
// join the patterns generated by the current item with all the other items of the fptree
multi_path_patterns.insert( curr_item_patterns.cbegin(), curr_item_patterns.cend() );
}
return multi_path_patterns;
}
}
