An efficient algorithm for mining quantitative association rules in large databases

Recently, there have been growing attentions on deducing association rules from large volume database systems. We introduce the problem of mining association rules in large relational tables containing quantitative attributes. In this paper, we propose a Large Interval Itemset generation method, which reflects locality when the attribute-value pairs are divided into sub-intervals. Since the method generates the itemset starting from high frequency area, we can produce more Large Interval Itemsets than other methods can do. This can result in better search performance and minimize the loss of characteristics of the original data. In the performance measurements conducted on the real-life dataset, such as the population census data, the proposed method out-performed the conventional ones.