A New Associative Classification Using Genetic Algorithm Pruned Decision Tree

Yun Huoy, Choo and Azah Kamilah, Muda (2013) A New Associative Classification Using Genetic Algorithm Pruned Decision Tree. Project Report. UTeM, Melaka, Malaysia. (Submitted)

PDF (Full Text) A_new_associative_classification_using_genetic_algorithm_pruned_decision_tree_Dr_Choo_Yun_Huoy.pdf - Submitted Version Restricted to Registered users only Download (6MB)

Abstract

Associative classification is known by its high accuracy and strong flexibility at handling unstructured data. However, it is still influenced by low quality dataset which consists of noise and irregular data. In this research, we proposed a new pruning technique to prune decision tree using genetic algorithm. In order to find the most optimum decision tree, decision tree is translated into fixed binary string that allowed going through evolution process. By using 5% of random mutation and 80% of Roulette crossover as common GA operators, we evolved each population. Finally the algorithm returns the chromosome with the best fitness value to the associative classification and the allele of the best chromosome will be used to prune the decision tree to produce a set of best rules. Our experiments on databases from UCI machine learning database repository show that the proposed associative classification is consistent, highly effective at classification of various categories of databases and has better average classification accuracy in comparison with CBA and CMAR.

Actions (login required)

View Item