NOV-RSI: A Novel Optimization Algorithm for Mining Rare Significance Itemsets

摘要

Rare itemsets mining is an important task for potential applications such as the detection of computer attacks, fraudulent transactions in financial institutions, bioinformatics and medicine. In the traditional data mining on transaction databases, such items have no weight (equal weight, as equal to 1). However, in real world application, each item often has a different weight (the importance/significance of each item). Therefore, we need to mine weighted frequent/rare itemsets on transaction databases. In this paper, we propose an algorithm for mining rare significance itemsets based on NOT satisfy the downward closure property. We propose an efficient algorithm called NOV-RSI. The experimental results show that the proposed algorithm performs faster than other existing algorithms on both real-life datasets of UCI and synthetic datasets generated by IBM Almaden.