Recently, multiple classifier systems (MCS) have been used for practical applications to improve classification accuracy. Self-generating neural networks (SGNN) are one of the suitable base-classifiers for MCS because of their simple setting and fast learning. However, the computational cost of the MCS increases in proportion to the number of SGNN. In this paper, we propose a novel pruning method for the structure of the SGNN in the MCS. Experiments have been conducted to compare the pruned MCS with an unpruned MCS, the MCS based on C4.5, and k-nearest neighbor method. The results show that the pruned MCS can improve its classification accuracy as well as reducing the computational cost. © Springer-Verlag Berlin Heidelberg 2003.
CITATION STYLE
Inoue, H., & Narihisa, H. (2003). Effective pruning method for a multiple classifier system based on self-generating neural networks. Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2714, 11–18. https://doi.org/10.1007/3-540-44989-2_2
Mendeley helps you to discover research relevant for your work.