Abstract:

Associative classification is a significant data mining technique. The schema with support and confidence is commonly employed in the stateoftheart associative classification methods. Since the classification based on support is very simple and the classification based on confidence fails to measure the correlation between itemset and class, these methods tend to generate many inferior rules. In this paper, we propose an improved associative classification approach based on support and enhancement ratio (ACSER). The ACSER considers the support of itemset both in the target class and in its complement class. Firstly, frequent enhancement ratio patterns are extracted from training data as candidate classification rules. Secondly, the ACSER ranks and prunes the extracted rules according to the rule intensity measured by confidence and enhancement ratio. Finally, the ACSER selects the best k rules to predict unknown objects. Experiments on 16 UCI datasets show that the improved approach has higher accuracy than the traditional approaches based on support and confidence.

Key words: data mining;associative classification;frequent itemset;rule intensity;classification accuracy