有向グラフ構造をもつ進化論的計算手法による相関ルールの抽出
スポンサーリンク
概要
- 論文の詳細を見る
Association rule mining is the discovery of association relationships or correlations among a set of attributes (items) in a database. Association rule in the form of 'If X then Y (X→Y)' is interpreted as 'database records that satisfy X are likely to satisfy Y.' Many techniques for association rule mining and associative classification have been proposed which have achieve quite effective performance. Recently, association rule mining tools using a graph structure based on the evolutionary computation technique have been proposed. The tools have been developed using a basic structure of Genetic Network Programming (GNP) and adopt a new strategy in evolution to execute tasks through generations. Association rules in the GNP-based methods are represented by the connections of GNP nodes. Extracted rules are accumulated in a rule library through GNP generations. Extended GNP-based rule mining methods can realize rule extraction from incomplete databases including missing values. Conventional rule mining methods cannot handle incomplete databases. In addition, algorithms capable of finding contrast rules showing different characteristics between two data sets, exception rule sets for focusing class and time related association rules have been proposed as extensions of the GNP-based method. It is not easy for the previous approaches to extract these rules. Experimental results using medical data showed the effectiveness and usefulness of the GNP-based rule extraction methods. GNP-based methods are suitable for the association rule mining from datasets in the medical field.
- 2011-12-31