Discovering Informative Association Rules for Associative Classification

The application of association rule mining to classification has led to a new family of classifiers which are often referred to as associative classifiers (ACs). An advantage of ACs is that they are rule-based and thus lend themselves to an easier interpretation. However, it is common knowledge that association rule mining typically yields a sheer number of rules defeating the purpose of a human readable model. Hence, selecting and ranking a small subset of high-quality rules without jeopardizing the classification accuracy is paramount but very challenging. In this paper, Entropy-AC, a new associative classifier based on entropy, is proposed. Information gain and informative rules are defined at first. Then, the algorithm for constructing associative classifier based on informative rules is presented. Experimental results show the proposed associative classifier is effective.

1. Introduction (heading 1)

A new classification approach known as associative classification integrates association mining and classification into a single system [1], [2]. Association mining, or pattern discovery, aims to discover descriptive knowledge from databases, while classification focuses on building a classification model for categorizing new data. Both association pattern discovery and classification rule mining are essential to practical data mining applications. If these two relevant jobs can be somehow integrated, great savings and conveniences to the user could result. Hence, considerable efforts have been made to integrate these two techniques into one system.