An Efficient Algorithm for Mining Frequent Itemsets

Several algorithms have been proposed so far to mine all the frequent itemsets in a transaction database. These algorithms differ from one another in the method of handling the candidate sets and the method of reducing the number of database passes. This paper thus attempts to propose a new data-mining algorithm for mining all the frequent itemsets in a transaction database. We present an algorithm, ¿FIMIT¿ which mines all the frequent itemsets in a transaction database using vertical transaction database format. The performance study shows that FIMIT is efficient and scalable over large databases, and is faster than the previously proposed methods. The main strength is not their speed (although they are not slow even outperforms Apriori), but the simplicity of their structure.