Mining fuzzy association rules on large numerical data: A data mining system for NAWN.

摘要

Mining numerical data has been and still is a burgeoning research area in computer science. This thesis introduces the problem of mining numerical data from very large databases, such as the data obtained from the Niagara Agricultural Weather Network (NAWN), using fuzzy logic and the data mining technique called association rule mining. The numerical data attributes are processed by converting them into categorical data or fuzzy sets using fuzzy logic. The converted numerical data can then be effectively mined using association rule mining algorithms. The first algorithm introduced is referred to as Fuzzy Apriori 1 (FA1), and is a simple implementation that allows each numeric datum to be represented by multiple fuzzy sets. This algorithm was found to take a lot of execution time. An improvement was made, resulting in a new algorithm referred to as Fuzzy Apriori 2 (FA2), which reduced the execution time by almost 60% but generated a lot of similar rules, compared to the rules generated by the FA1 algorithm. (Abstract shortened by UMI.)