/**
* Run the algorithm.
*
* @param minSupport Minsup as a percentage (ex: 0.05 = 5 %)
* @param minConfidence minimum confidence (a value between 0 and 1).
* @param input the input file path
* @param output the output file path
* @param windowSize a window size
* @throws IOException exception if there is an error reading/writing files
*/
public void runAlgorithm(
double minSupport, double minConfidence, String input, String output, int windowSize)
throws IOException {
// load the input file into memory
try {
this.database = new SequenceDatabase();
database.loadFile(input);
} catch (Exception e) {
e.printStackTrace();
}
// convert minimum support to a relative minimum support (integer)
this.minsuppRelative = (int) Math.ceil(minSupport * database.size());
// run the algorithm
runAlgorithm(input, output, minsuppRelative, minConfidence, windowSize);
}
/**
* Run the algorithm.
*
* @param relativeMinSupport minsup as a a relative value (integer)
* @param minConfidence minimum confidence (a value between 0 and 1).
* @param input the input file path
* @param output the output file path
* @param windowSize a window size
* @throws IOException exception if there is an error reading/writing files
*/
public void runAlgorithm(
String input, String output, int relativeMinSupport, double minConfidence, int windowSize)
throws IOException {
this.minconf = minConfidence;
// read the database into memory
if (database == null) {
try {
this.database = new SequenceDatabase();
database.loadFile(input);
} catch (Exception e) {
e.printStackTrace();
}
}
// IMPORTANT : THIS IS A FIX SO THAT THE DEFINITION IS THE SAME AS IN THE ARTICLE!!
this.windowSize = windowSize + 1;
// if minsup is 0, set it to 1
this.minsuppRelative = relativeMinSupport;
if (this.minsuppRelative == 0) { // protection
this.minsuppRelative = 1;
}
// reset the stats for memory usage
MemoryLogger.getInstance().reset();
// prepare the object for writing the output file
writer = new BufferedWriter(new FileWriter(output));
// save the start time
timeStart = System.currentTimeMillis(); // for stats
// remove infrequent items from the database
removeItemsThatAreNotFrequent(database);
// note frequent items in a list "listFrequents"
List<String> listFrequents = new ArrayList<String>();
// for each item
for (Entry<String, Map<Integer, Occurence>> entry : mapItemCount.entrySet()) {
// if it is frequent
if (entry.getValue().size() >= minsuppRelative) {
// add the item to the list
listFrequents.add(entry.getKey());
}
}
// FOR EACH FREQUENT ITEM WE COMPARE WITH EACH OTHER FREQUENT ITEM TO
// TRY TO GENERATE A RULE 1-1.
for (int i = 0; i < listFrequents.size(); i++) {
String intI = listFrequents.get(i);
Map<Integer, Occurence> occurencesI = mapItemCount.get(intI);
for (int j = i + 1; j < listFrequents.size(); j++) {
String intJ = listFrequents.get(j);
Map<Integer, Occurence> occurencesJ = mapItemCount.get(intJ);
// (1) Calculate tidsI, tidsJ, tidsJ-->J and tidsI->J
Set<Integer> tidsI = new HashSet<Integer>();
Set<Integer> tidsJ = null;
Set<Integer> tidsIJ = new HashSet<Integer>();
Set<Integer> tidsJI = new HashSet<Integer>();
// for each occurence of I
looptid:
for (Occurence occI : occurencesI.values()) {
// add the sequenceID to tidsI
tidsI.add(occI.sequenceID);
// if J does not appear in that sequence continue loop
Occurence occJ = occurencesJ.get(occI.sequenceID);
if (occJ == null) {
continue looptid;
}
// make a big loop to compare if I appears before
// J in that sequence and
// if J appears before I
boolean addedIJ = false;
boolean addedJI = false;
// for each occurence of I in that sequence
loopIJ:
for (Short posI : occI.occurences) {
// for each occurence of J in that sequence
for (Short posJ : occJ.occurences) {
if (!posI.equals(posJ) && Math.abs(posI - posJ) <= windowSize) {
if (posI <= posJ) {
// if I is before J
tidsIJ.add(occI.sequenceID);
addedIJ = true;
} else {
// if J is before I
tidsJI.add(occI.sequenceID);
addedJI = true;
}
// if we have found that I is before J and J is before I
// we don't need to continue.
if (addedIJ && addedJI) {
break loopIJ;
}
}
}
}
}
// END
// (2) check if the two itemsets have enough common tids
// if not, we don't need to generate a rule for them.
// create rule IJ
if (tidsIJ.size() >= minsuppRelative) {
// calculate the confidence of I ==> J
double confIJ = ((double) tidsIJ.size()) / occurencesI.size();
// create itemset of the rule I ==> J
String[] itemset1 = new String[] {intI};
String[] itemset2 = new String[] {intJ};
// if the confidence is high enough, save the rule
if (confIJ >= minConfidence) {
saveRule(tidsIJ, confIJ, itemset1, itemset2);
}
// Calculate tidsJ.
tidsJ = new HashSet<Integer>();
for (Occurence occJ : occurencesJ.values()) {
tidsJ.add(occJ.sequenceID);
}
// recursive call to try to expand the rule
expandLeft(itemset1, itemset2, tidsI, tidsIJ);
expandRight(itemset1, itemset2, tidsI, tidsJ, tidsIJ);
}
// create rule JI
if (tidsJI.size() >= minsuppRelative) {
double confJI = ((double) tidsJI.size()) / occurencesJ.size();
// create itemsets for that rule
String[] itemset1 = new String[] {intI};
String[] itemset2 = new String[] {intJ};
// if the rule has enough confidence, save it!
if (confJI >= minConfidence) {
saveRule(tidsJI, confJI, itemset2, itemset1);
// rules.addRule(ruleJI);
}
// Calculate tidsJ.
if (tidsJ == null) {
tidsJ = new HashSet<Integer>();
for (Occurence occJ : occurencesJ.values()) {
tidsJ.add(occJ.sequenceID);
}
}
// recursive call to try to expand the rule
expandRight(itemset2, itemset1, tidsJ, tidsI, tidsJI /*, occurencesJ, occurencesI*/);
expandLeft(itemset2, itemset1, tidsJ, tidsJI /*, occurencesI*/);
}
}
}
// save the end time for the execution of the algorithm
timeEnd = System.currentTimeMillis(); // for stats
// close the file
writer.close();
database = null;
}