/**
* This methods checks if a seq. pattern "pattern2" is strictly contained in a seq. pattern
* "pattern1".
*
* @param pattern1 a sequential pattern
* @param pattern2 another sequential pattern
* @return true if the pattern1 contains pattern2.
*/
boolean strictlyContains(PrefixVMSP pattern1, PrefixVMSP pattern2) {
// // if pattern2 is larger or equal in size, then it cannot be contained in pattern1
// if(pattern1.size() <= pattern2.size()){
// return false;
// }
// To see if pattern2 is strictly contained in pattern1,
// we will search for each itemset i of pattern2 in pattern1 by advancing
// in pattern 1 one itemset at a time.
int i = 0; // position in pattern2
int j = 0; // position in pattern1
while (true) {
// if the itemset at current position in pattern1 contains the itemset
// at current position in pattern2
if (pattern1.get(j).containsAll(pattern2.get(i))) {
// go to next itemset in pattern2
i++;
// if we reached the end of pattern2, then return true
if (i == pattern2.size()) {
return true;
}
}
// go to next itemset in pattern1
j++;
// if we reached the end of pattern1, then pattern2 is not strictly included
// in it, and return false
if (j >= pattern1.size()) {
return false;
}
// // lastly, for optimization, we check how many itemsets are left to be matched.
// // if there is less itemsets left in pattern1 than in pattern2, then it will
// // be impossible to get a total match, and so we return false.
if ((pattern1.size() - j) < pattern2.size() - i) {
return false;
}
}
}
/**
* This is the dfsPruning method as described in the SPAM paper.
*
* @param prefix the current prefix
* @param prefixBitmap the bitmap corresponding to the current prefix
* @param sn a list of items to be considered for i-steps
* @param in a list of items to be considered for s-steps
* @param hasToBeGreaterThanForIStep
* @param m size of the current prefix in terms of items
* @param lastAppendedItem the last appended item to the prefix
* @throws IOException if there is an error writing a pattern to the output file
* @return TRUE IF A FREQUENT PATTERN WAS CREATED USING THE PREFIX.
*/
boolean dfsPruning(
PrefixVMSP prefix,
Bitmap prefixBitmap,
List<Integer> sn,
List<Integer> in,
int hasToBeGreaterThanForIStep,
int m,
Integer lastAppendedItem)
throws IOException {
boolean atLeastOneFrequentExtension = false;
// System.out.println(prefix.toString());
// ====== S-STEPS ======
// Temporary variables (as described in the paper)
List<Integer> sTemp = new ArrayList<Integer>();
List<Bitmap> sTempBitmaps = new ArrayList<Bitmap>();
// for CMAP pruning, we will only check against the last appended item
Map<Integer, Integer> mapSupportItemsAfter = coocMapAfter.get(lastAppendedItem);
// for each item in sn
loopi:
for (Integer i : sn) {
// LAST POSITION PRUNING
/*if (useLastPositionPruning && lastItemPositionMap.get(i) < prefixBitmap.firstItemsetID) {
// System.out.println("TEST");
continue loopi;
}*/
// CMAP PRUNING
// we only check with the last appended item
if (useCMAPPruning) {
if (mapSupportItemsAfter == null) {
continue loopi;
}
Integer support = mapSupportItemsAfter.get(i);
if (support == null || support < minsup) {
// System.out.println("PRUNE");
continue loopi;
}
}
// perform the S-STEP with that item to get a new bitmap
Bitmap.INTERSECTION_COUNT++;
Bitmap newBitmap =
prefixBitmap.createNewBitmapSStep(verticalDB.get(i), sequencesSize, lastBitIndex, maxGap);
// if the support is higher than minsup
if (newBitmap.getSupportWithoutGapTotal() >= minsup) {
// record that item and pattern in temporary variables
sTemp.add(i);
sTempBitmaps.add(newBitmap);
}
}
// for each pattern recorded for the s-step
for (int k = 0; k < sTemp.size(); k++) {
// STRATEGY: NEWWW
atLeastOneFrequentExtension = true;
int item = sTemp.get(k);
// create the new prefix
PrefixVMSP prefixSStep = prefix.cloneSequence();
prefixSStep.addItemset(new Itemset(item));
if (item % 2 == 0) {
prefixSStep.sumOfEvenItems = item + prefix.sumOfEvenItems;
prefixSStep.sumOfOddItems = prefix.sumOfOddItems;
} else {
prefixSStep.sumOfEvenItems = prefix.sumOfEvenItems;
prefixSStep.sumOfOddItems = item + prefix.sumOfOddItems;
}
// prefixSStep.sumOfItems = item + prefix.sumOfItems;
// create the new bitmap
Bitmap newBitmap = sTempBitmaps.get(k);
// save the pattern to the file
if (newBitmap.getSupport() >= minsup) {
boolean hasFrequentExtension = false;
// recursively try to extend that pattern
if (maximumPatternLength > m) {
hasFrequentExtension =
dfsPruning(prefixSStep, newBitmap, sTemp, sTemp, item, m + 1, item);
}
if (hasFrequentExtension == false) {
savePatternMultipleItems(prefixSStep, newBitmap, m);
}
}
}
Map<Integer, Integer> mapSupportItemsEquals = coocMapEquals.get(lastAppendedItem);
// ======== I STEPS =======
// Temporary variables
List<Integer> iTemp = new ArrayList<Integer>();
List<Bitmap> iTempBitmaps = new ArrayList<Bitmap>();
// for each item in in
loop2:
for (Integer i : in) {
// the item has to be greater than the largest item
// already in the last itemset of prefix.
if (i > hasToBeGreaterThanForIStep) {
// LAST POSITION PRUNING
/*if (useLastPositionPruning && lastItemPositionMap.get(i) < prefixBitmap.firstItemsetID) {
continue loop2;
}*/
// CMAP PRUNING
if (useCMAPPruning) {
if (mapSupportItemsEquals == null) {
continue loop2;
}
Integer support = mapSupportItemsEquals.get(i);
if (support == null || support < minsup) {
continue loop2;
}
}
// Perform an i-step with this item and the current prefix.
// This creates a new bitmap
Bitmap.INTERSECTION_COUNT++;
Bitmap newBitmap =
prefixBitmap.createNewBitmapIStep(verticalDB.get(i), sequencesSize, lastBitIndex);
// If the support is no less than minsup
if (newBitmap.getSupport() >= minsup) {
// record that item and pattern in temporary variables
iTemp.add(i);
iTempBitmaps.add(newBitmap);
}
}
}
// for each pattern recorded for the i-step
for (int k = 0; k < iTemp.size(); k++) { // STRATEGY: NEWWW
atLeastOneFrequentExtension = true;
int item = iTemp.get(k);
// create the new prefix
PrefixVMSP prefixIStep = prefix.cloneSequence();
prefixIStep.getItemsets().get(prefixIStep.size() - 1).addItem(item);
if (item % 2 == 0) {
prefixIStep.sumOfEvenItems = item + prefix.sumOfEvenItems;
prefixIStep.sumOfOddItems = prefix.sumOfOddItems;
} else {
prefixIStep.sumOfEvenItems = prefix.sumOfEvenItems;
prefixIStep.sumOfOddItems = item + prefix.sumOfOddItems;
}
// create the new bitmap
Bitmap newBitmap = iTempBitmaps.get(k);
// recursively try to extend that pattern
boolean hasFrequentExtension = false;
if (maximumPatternLength > m) {
hasFrequentExtension = dfsPruning(prefixIStep, newBitmap, sTemp, iTemp, item, m + 1, item);
}
if (hasFrequentExtension == false) {
// save the pattern
savePatternMultipleItems(prefixIStep, newBitmap, m);
}
}
// check the memory usage
MemoryLogger.getInstance().checkMemory();
return atLeastOneFrequentExtension || useStrategyForwardExtensionChecking == false;
}