Mining Association Rules from a Dynamic Probabilistic Numerical Dataset Using Estimated-Frequent Uncertain-Itemsets

Abstract

In recent years, many new applications, such as location-based services, sensor monitoring systems, and data integration, have shown a growing amount of importance of uncertain data mining. In addition, due to instrument errors, imprecise of sensor monitoring systems, and so on, real-world data tend to be numerical data with inherent uncertainty. Thus, mining association rules from an uncertain, especially probabilistic numerical dataset has been studied recently. However, a probabilistic numerical dataset often grows as new data append. Thus, developing a mining algorithm that can incrementally maintain discovered information is quite important. In this paper, we have designed an efficient, incremental mining algorithm to mine association rules from a probabilistic numeric dataset using estimated-frequent uncertain-itemsets. By using a user-specified support threshold, estimated-frequent uncertain-itemsets could act as a gap to avoid small itemsets becoming large in the updated dataset when new transactions are inserted. As a result, the algorithm has execution time faster than that of previous methods. An illustrated example is given to demonstrate the procedures of the algorithm.