Abstract

The original purpose of data mining is for analysis of supermarket transaction data. Now with the rapid development in business, industry and science, data mining is used in lots of domains, so mining interesting information from large database becomes more important. Data mining includes two main parts: frequent itemsets mining and association rules mining. And frequent itemsets mining plays an essential role between them.
Our thesis is focused on frequent itemsets mining. Previous studies on frequent itemsets mining is at single or multiple concept level, however, mining frequent itemsets at flexible multiple concept level may help finding more specific and useful information from huge data. In this thesis, four methods are introduced for mining frequent itemsets at flexible multiple level by extension of Apriori and Eclat algorithms. We also implement two algorithms for frequent pairs mining. We draw some conclusions about which method is suitable for which distributions of data.