/**
* This method creates a copy of the sequence and add a given item to the last itemset of the
* sequence. It sets the support of the sequence as the support of the item.
*
* @param prefix the sequence
* @param item the item
* @return the new sequence
*/
private SequentialPattern appendItemToPrefixOfSequence(SequentialPattern prefix, Integer item) {
SequentialPattern newPrefix = prefix.cloneSequence();
// add to the last itemset
Itemset itemset = newPrefix.get(newPrefix.size() - 1);
itemset.addItem(item);
return newPrefix;
}
/**
* This method saves a sequential pattern to the output file or in memory, depending on if the
* user provided an output file path or not when he launched the algorithm
*
* @param prefix the pattern to be saved.
* @throws IOException exception if error while writing the output file.
*/
private void savePattern(SequentialPattern prefix) throws IOException {
// increase the number of pattern found for statistics purposes
patternCount++;
// if the result should be saved to a file
if (writer != null) {
StringBuffer r = new StringBuffer("");
for (Itemset itemset : prefix.getItemsets()) {
// r.append('(');
for (String item : itemset.getItems()) {
String string = item.toString();
r.append(string);
r.append(' ');
}
r.append("-1 ");
}
//
// // print the list of Pattern IDs that contains this pattern.
// if(prefix.getSequencesID() != null){
// r.append("SID: ");
// for(Integer id : prefix.getSequencesID()){
// r.append(id);
// r.append(' ');
// }
// }
r.append(" #SUP: ");
r.append(prefix.getSequencesID().size());
writer.write(r.toString());
writer.newLine();
} // otherwise the result is kept into memory
else {
patterns.addSequence(prefix, prefix.size());
}
}
/**
* This is the "backscan-pruning" strategy described in the BIDE+ paper to avoid extending some
* prefixs that are guaranteed to not generate a closed pattern (see the BIDE+ paper for details).
*
* @param prefix the current prefix
* @param projectedContext the projected database
* @return boolean true if we should not extend the prefix
*/
private boolean checkBackScanPruning(
SequentialPattern prefix, List<PseudoSequenceBIDE> projectedContext) {
// DEBUGGIN
if (prefix.size() == 1 && prefix.get(0).get(0) == 5) {
System.out.println("PREFIX 5 ");
}
//
// See the BIDE+ paper for details about this method.
// For the number of item occurences that can be generated with this prefix:
for (int i = 0; i < prefix.getItemOccurencesTotalCount(); i++) {
Set<Integer> alreadyVisitedSID = new HashSet<Integer>();
// Create a Map of pairs to count the support of items (represented by a pair)
// in the ith semi-maximum periods
Map<PairBIDE, PairBIDE> mapPaires = new HashMap<PairBIDE, PairBIDE>();
// SOME CODE USED BY "findAllFrequentPairsForBackwardExtensionCheck"
Integer itemI = prefix.getIthItem(i); // iPeriod
Integer itemIm1 = null; // iPeriod -1
if (i > 0) {
itemIm1 = prefix.getIthItem(i - 1);
}
// // END NEW
// (1) For each i, we build the list of maximum periods
// for each sequence in the original database
for (PseudoSequenceBIDE pseudoSequence : projectedContext) {
int sequenceID = pseudoSequence.sequence.getId();
alreadyVisitedSID.add(pseudoSequence.sequence.getId());
Position currentCutPosition =
new Position(pseudoSequence.firstItemset, pseudoSequence.firstItem);
PseudoSequenceBIDE sequence = initialDatabase.get(sequenceID);
PseudoSequenceBIDE period =
sequence.getIthSemiMaximumPeriodOfAPrefix(prefix.getItemsets(), i, currentCutPosition);
if (period != null) {
// // we add it to the list of maximum periods
boolean hasExtension =
findAllFrequentPairsForBackwardExtensionCheck(
prefix.getAbsoluteSupport(),
prefix,
period,
i,
mapPaires,
itemI,
itemIm1,
currentCutPosition);
if (hasExtension) {
return true;
}
}
}
}
return false;
}
/**
* This method creates a copy of the sequence and add a given item to the last itemset of the
* sequence. It sets the support of the sequence as the support of the item.
*
* @param prefix the sequence
* @param item the item
* @return the new sequence
*/
private SequentialPattern appendItemToPrefixOfSequence(SequentialPattern prefix, String item) {
SequentialPattern newPrefix = prefix.cloneSequence();
Itemset itemset = newPrefix.get(newPrefix.size() - 1);
itemset.addItem(item);
return newPrefix;
}
/**
* Method to recursively grow a given sequential pattern.
*
* @param prefix the current sequential pattern that we want to try to grow
* @param database the current projected sequence database
* @throws IOException exception if there is an error writing to the output file
*/
private int recursion(SequentialPattern prefix, List<PseudoSequenceBIDE> contexte)
throws IOException {
if (prefix.size() == 1 && prefix.get(0).get(0) == 5) {
System.out.println("PREFIXBY5 : " + prefix);
}
// find frequent items of size 1 in the current projected database.
Set<PairBIDE> pairs = findAllFrequentPairs(prefix, contexte);
// DEBUGING
System.out.println("PREFIX: " + prefix);
if (prefix.size() == 2
&& prefix.get(0).size() == 3
&& prefix.get(1).size() == 3
&& prefix.get(0).get(0) == 1
&& prefix.get(0).get(1) == 3
&& prefix.get(0).get(2) == 5
&& prefix.get(1).get(0) == 2
&& prefix.get(1).get(1) == 3
&& prefix.get(1).get(2) == 5) {
System.out.println("DEBUGGING !!!!!!!! #### (1 3 5 )(2 3 5 )");
}
if (prefix.size() == 2
&& prefix.get(0).size() == 3
&& prefix.get(1).size() == 2
&& prefix.get(0).get(0) == 1
&& prefix.get(0).get(1) == 3
&& prefix.get(0).get(2) == 5
&& prefix.get(1).get(0) == 2
&& prefix.get(1).get(1) == 3) {
System.out.println("DEBUGGING !!!!!!!! #### (1 3 5 )(2 3 )");
}
// we will keep tract of the maximum support of patterns
// that can be found with this prefix, to check
// for forward extension when this method returns.
int maxSupport = 0;
// For each pair found (a pair is an item with a boolean indicating if it
// appears in an itemset that is cut (a postfix) or not, and the sequence IDs
// where it appears in the projected database).
for (PairBIDE pair : pairs) {
// if the item is frequent.
if (pair.getCount() >= minsuppAbsolute) {
// create the new postfix by appending this item to the prefix
SequentialPattern newPrefix;
// if the item is part of a postfix
if (pair.isPostfix()) {
// we append it to the last itemset of the prefix
newPrefix = appendItemToPrefixOfSequence(prefix, pair.getItem()); // is =<is, (deltaT,i)>
} else { // else, we append it as a new itemset to the sequence
newPrefix = appendItemToSequence(prefix, pair.getItem());
}
// build the projected database with this item
// long start = System.currentTimeMillis();
List<PseudoSequenceBIDE> projectedContext =
buildProjectedDatabase(
pair.getItem(), contexte, pair.isPostfix(), pair.getSequenceIDs());
// debugProjectDBTime += System.currentTimeMillis() - start;
// create new prefix
newPrefix.setSequenceIDs(pair.getSequenceIDs());
// variable to keep track of the maximum support of extension
// with this item and this prefix
if (projectedContext.size() >= minsuppAbsolute) {
int maxSupportOfSuccessors = 0;
if (!checkBackScanPruning(newPrefix, projectedContext)) {
maxSupportOfSuccessors = recursion(newPrefix, projectedContext); // r�cursion
}
// check the forward extension for the prefix
// if no forward extension
if (newPrefix.getSequenceIDs().size() != maxSupportOfSuccessors) {
// if there is no backward extension
if (!checkBackwardExtension(newPrefix, projectedContext)) {
// save the pattern
savePattern(newPrefix);
}
}
} else {
if (!checkBackwardExtension(newPrefix, projectedContext)) {
// save the pattern
savePattern(newPrefix);
}
}
// record the largest support of patterns found starting
// with this prefix until now
if (newPrefix.getAbsoluteSupport() > maxSupport) {
maxSupport = newPrefix.getAbsoluteSupport();
}
}
}
return maxSupport; // return the maximum support generated by extension of the prefix
}