An Accurate Fuzzy Frequent Pattern Based Classifier Using Confidence Tuning

Associative classifier algorithms combine two data mining paradigms, namely sample classification andassociation rule mining. These methods are very interesting for building an accurate classification model in a wide area of real-world applications. Lately, many methods have been presented to integrate associative classifiers with fuzzy set theory, in order to improve the quality of previous algorithms. This paper presents a three-step fuzzy frequent pattern (FFP) based classifier which uses an Apriori like algorithm to generate a large number of FFPs from each data class. Our algorithm in the second stepselects a subset of useful FFPs and removes redundant ones. Finally, in order to tune the boundaries between various data classes, we use a confidence improvement process. We tested our algorithm on six real-world datasets and compared the achieved results with two well-known fuzzy associative classifier algorithms.

An Accurate Fuzzy Frequent Pattern Based Classifier Using Confidence Tuning Keywords:

Fuzzy associative classifier , Certainty factor tuning , Fuzzy frequent pattern , Fuzzy rule