CONSTRUCTION OF CLUSTERING AND CLASSIFICATION MODELS BY INTEGRATING FUZZY ART, CART AND NEURAL NETWORK APPROACHES

Abstract

ABSTRACT With the help of advanced information technology, it is no longer a difficult task in collecting relevant data sets of customers. However, the data sets growth too fast, it is not easy to identify the complicate relationship in the huge data sets. Moreover, the traditional management information systems can only conduct basic descriptive statistics with respect to the collected data and therefore unable to dig out important and latent information inside the data. Data mining is a fast growing application area in business. With data mining techniques, it allows the possibility of computer-driven exploration of the data, and we don't need to assume some hypothesis for the data. The purpose of this research is to provide a complete data analysis process, and there are two main stages included. In the first stage, we used Fuzzy ART to identify an appropriate number of clusters for the data. In the second stage, we integrated neural networks and classification and regression tree (CART) to solve the classification problems. To demonstrate the efficiency of the proposal approaches, classification tasks are performed on two data sets, the Zoo data (adapted from UCI Machine Learning Repository) and one simulated data. As the results reveal, the proposed integrated approach provides a better initial solution than the conventional neural networks. Besides, comparing with the pure neural network approach, the classification accuracies increase for both cases in the proposed methodology.