Discovering patterns based on fuzzy logic theory

Abstract

This study investigates the formulation of fuzzy logic as integrated component of the proposed model in data mining in order to classify the dataset prior to the implementation of data mining tools such summarization, association rule discovery, and prediction. The novel contribution of this paper is the fuzzification of the dataset prior to pattern discovery. The model is compared to the classical clustering, regression model, and neural network using the Internet usage database available at the UCI Knowledge Discovery on Databases (KDD) archive. Our test is anchored on parameters like relevant measure, processing performance, discovered rules or patterns and practical use of the findings. The proposed model indicates adequate performance in clustering, higher clustering accuracy and efficient pattern discovery compared with the other models. © Springer-Verlag Berlin Heidelberg 2006.