2009 IEEE International Conference on
Systems, Man, and Cybernetics |
 |
- [P219]
Fuzzy Classification Rule Minining Based on Genetic Network Programming Algorithm.
Author(s): Karla Taboada, Shingo Mabu, Eloy Gonzales, Shimada Kaoru, Kotaro HirasawaAbstract
Association rule-based classification is one of the most important data mining techniques applied to many scientific problems. In the last few years, extensive research has been carried out to develop enhanced methods and obtained higher classification accuracies than traditional classifiers. However, the current studies show that the association rule-based classifiers may also suffer some problems inherited from association rule mining such as handling of (1) continuous data, (2) the support/confidence framework, and (3) a large set of discovered rules. In order to deal with the above problems, a novel fuzzy classification model has been proposed in this paper that uses fuzzy class association rules as knowledge representation based on Genetic Network Programming (GNP). GNP is one of the evolutionary optimization algorithms that uses directed graph structures as solutions instead of strings (Genetic Algorithms) or trees (Genetic Programming). Therefore, GNP can deal with more complex problems by using the higher expression ability of graph structures. Our model consists of two major phases: 1) generating fuzzy class association rules by using GNP, 2) building a classifier based on the extracted fuzzy rules. In the first phase, the task is to extract fuzzy class association rules from a fuzzified training set using a GNP-based algorithm. Moreover, fuzzy rules are updated by evolving the membership functions through generations in order to obtain the best qualified rules for building the classifier. In the second phase, all of the generated fuzzy rules are used to predict the class of the test set. For each test data, the classifier computes the average matching degree between the data and the rules in each class. Finally, the class with the largest matching degree is assigned to the test data. The performance of our algorithm has been compared with other relevant algorithms and the experimental results show the advantages and effectiveness of the proposed model.