An Evolutionary Approach to Data Classification - Hybrid Real-Coded Genetic Algorithm with Pruning

We have already proposed a hybrid real-coded genetic algorithm with local search (HRGA/LS) for improving the search performance of a conventional real-coded genetic algorithm. To further enhance the generalization ability of classification models by HRGA/LS, this paper proposes a hybrid real-coded genetic algorithm with pruning (HRGA/P). A crucial idea here is the introduction of a regularizer into a fitness function for better generalization. Accordingly, the proposed algorithm has the following advantages: 1) finding near optimal classification models efficiently by a hybrid technique, 2) improving the generalization ability of classification models by a regularization technique. Applications of the proposed algorithm to an iris classification problem well demonstrate its effectiveness. Our experimental results clearly indicate that HRGA/P has higher classification performance not only in training data but also in test data (classification rate: 96.6%) than the conventional algorithms such as backpropagation (classification rate: 94.1%) and structural learning with forgetting (classification rate: 95.0%).