/**
* 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) {
StringBuilder r = new StringBuilder("");
for (Itemset itemset : prefix.getItemsets()) {
// r.append('(');
for (Integer 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.getSequenceIDs().size());
if (showSequenceIdentifiers) {
r.append(" #SID: ");
for (Integer sid : prefix.getSequenceIDs()) {
r.append(sid);
r.append(" ");
}
}
writer.write(r.toString());
writer.newLine();
} // otherwise the result is kept into memory
else {
patterns.addSequence(prefix, prefix.size());
}
}
/**
* 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 {
// find frequent items of size 1 in the current projected database.
Set<PairBIDE> pairs = findAllFrequentPairs(prefix, contexte);
// 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, pair.getSequenceIDs())) {
maxSupportOfSuccessors = recursion(newPrefix, projectedContext); // récursion
}
// 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, pair.getSequenceIDs())) {
// save the pattern
savePattern(newPrefix);
}
}
} else {
if (!checkBackwardExtension(newPrefix, pair.getSequenceIDs())) {
// 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
}
/**
* Method to update the support count of item in a maximum period
*
* @param prefix the current prefix
* @param mapPaires
* @param maximum periods a maximum period
* @return a set of pairs indicating the support of items (note that a pair distinguish between
* items in a postfix, prefix...).
*/
protected boolean findAllFrequentPairsForBackwardExtensionCheck(
int seqProcessedCount,
SequentialPattern prefix,
PseudoSequenceBIDE maximumPeriod,
int iPeriod,
Map<PairBIDE, PairBIDE> mapPaires,
Integer itemI,
Integer itemIm1) {
int supportToMatch = prefix.getSequenceIDs().size();
int maxPeriodSize = maximumPeriod.size();
// for each itemset in that period
for (int i = 0; i < maxPeriodSize; i++) {
int sizeOfItemsetAtI = maximumPeriod.getSizeOfItemsetAt(i);
// NEW
boolean sawI = false; // sawI after current position
boolean sawIm1 = false; // sawI-1 before current position
// END NEW
// NEW march 20 2010 : check if I is after current position in current itemset
for (int j = 0; j < sizeOfItemsetAtI; j++) {
Integer item = maximumPeriod.getItemAtInItemsetAt(j, i);
if (item.equals(itemI)) {
sawI = true;
} else if (item > itemI) {
break;
}
}
// END NEW
for (int j = 0; j < sizeOfItemsetAtI; j++) {
Integer item = maximumPeriod.getItemAtInItemsetAt(j, i);
if (itemIm1 != null && item == itemIm1) {
sawIm1 = true;
}
boolean isPrefix = maximumPeriod.isCutAtRight(i);
boolean isPostfix = maximumPeriod.isPostfix(i);
// END NEW
PairBIDE paire = new PairBIDE(isPrefix, isPostfix, item);
if (seqProcessedCount >= minsuppAbsolute) {
// $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
// normal case
if (addPair(mapPaires, maximumPeriod.getId(), paire, supportToMatch)) {
return true;
}
// NEW: special cases
if (sawIm1) {
PairBIDE paire2 = new PairBIDE(isPrefix, !isPostfix, item);
if (addPair(mapPaires, maximumPeriod.getId(), paire2, supportToMatch)) {
return true;
}
}
if (sawI) {
PairBIDE paire2 = new PairBIDE(!isPrefix, isPostfix, item);
if (addPair(mapPaires, maximumPeriod.getId(), paire2, supportToMatch)) {
return true;
}
}
// END NEW
} else { // $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
// normal case
addPairWithoutCheck(mapPaires, maximumPeriod.getId(), paire);
// // NEW: special cases
if (sawIm1) {
PairBIDE paire2 = new PairBIDE(isPrefix, !isPostfix, item);
addPairWithoutCheck(mapPaires, maximumPeriod.getId(), paire2);
}
if (sawI) {
PairBIDE paire2 = new PairBIDE(!isPrefix, isPostfix, item);
addPairWithoutCheck(mapPaires, maximumPeriod.getId(), paire2);
}
// END NEW
}
}
}
return false;
}