// The Apriori Algorithm
private static List<Itemset> apriori(List<Transaction> transactions, double minsup) {
if (transactions.isEmpty()) {
return null;
}
// Generate frequent 1-itemset
List<Integer> occurancesOfItems = new ArrayList<>();
for (int i = 0; i != transactions.get(0).getNumOfTotalItems(); ++i) {
occurancesOfItems.add(0);
}
for (Transaction transaction : transactions) {
for (int index : transaction.items()) {
occurancesOfItems.set(index, occurancesOfItems.get(index) + 1);
}
}
int numOfTransactions = transactions.size();
List<Itemset> frequentOneItemsets = new ArrayList<>();
for (int i = 0; i != occurancesOfItems.size(); ++i) {
if ((double) occurancesOfItems.get(i) / numOfTransactions >= minsup) {
Itemset itemset = new Itemset(i);
frequentOneItemsets.add(itemset);
}
}
// A list of itemsets to store the frequent k-itemsets, k = 1, 2, ...
List<List<Itemset>> listOfFrequentItemsets = new ArrayList<List<Itemset>>();
listOfFrequentItemsets.add(
new ArrayList<Itemset>()); // case: k = 0, we fill it with an empty itemsets
listOfFrequentItemsets.add(frequentOneItemsets); // case: k = 1
List<Itemset> prunedCandidates;
List<Itemset> frequentItemsets;
int k = 1;
while ((frequentItemsets = listOfFrequentItemsets.get(k)) != null) {
List<Itemset> newCandidates = generateCandidates(frequentItemsets);
prunedCandidates = prune(newCandidates, frequentItemsets);
List<Itemset> candidates = determineFrequentItemsets(prunedCandidates, transactions, minsup);
listOfFrequentItemsets.add(candidates);
k++;
}
// Union the frequent itemsets as result
List<Itemset> result = new ArrayList<>();
for (List<Itemset> list : listOfFrequentItemsets) {
if (list != null) {
result.addAll(list);
}
}
return result;
}
