Mining maximal frequent itemsets: A java implementation of FPMAX algorithm

Mining maximal frequent itemsets is an important issue in many data mining applications. In our thesis work on selection and tuning of indices in data warehouses, we have proposed a strategy based on mining maximal frequent itemsets in order to determine a set of candidate indices from a given workload. In a first step we have to select an algorithm, for mining maximal frequent itemsets, to implement. Experimental results in the repository of the workshops on Frequent Itemset Mining Implementations (http://fimi.cs.helsinki.fi/), shows that FPMAX has the best performance. Therefore, we have selected it for our own implementation in java language. FPMAX is an extension of FP-Growth method for mining maximal frequent itemsets only. We tested our implementation on two benchmark databases MUSHROOM and RETAIL. We compare our results with the best implementations available in the repository mentioned earlier. Our implementation showed good performances compared with the others. However, the comparison of response times published in FIMI 2004, for the chosen implementations, could not be replicated.