Genetic Algorithms in Search, Optimization and Machine Learning
Genetic Algorithms in Search, Optimization and Machine Learning
Fuzzy Sets and Systems: Theory and Applications
Fuzzy Sets and Systems: Theory and Applications
Truth from Trash: How Learning Makes Sense
Truth from Trash: How Learning Makes Sense
Pattern Recognition and Machine Learning (Information Science and Statistics)
Pattern Recognition and Machine Learning (Information Science and Statistics)
Selecting features in microarray classification using ROC curves
Pattern Recognition
Differentially Expressed Gene Identification Based on Separability Index
ICMLA '09 Proceedings of the 2009 International Conference on Machine Learning and Applications
IEEE Transactions on Fuzzy Systems
Hi-index | 0.00 |
The identification of class differentiating genes is central to microarray data classification. Genes are ranked in order of differential expression and the optimal top ranking genes are selected as features for classification. In this paper, a new approach to gene ranking, based on a fuzzy inference system - the Fuzzy Gene Filter - is presented and compared to classical ranking approaches (the t-test, Wilcoxon test and ROC analysis). Two performance metrics are used; maximum Separability Index and highest cross-validation accuracy. The techniques were implemented on two publically available data-sets. The Fuzzy Gene Filter outperformed the other techniques both with regards to maximum Separability Index, as well as highest cross-validation accuracy. For the prostate data-set it a attained a Leave-one-out cross-validation accuracy of 96.1% and for the lymphoma data-set, 100%. The Fuzzy Gene Filter cross-validation accuracies were also higher than those recorded in previous publications which used the same data-sets. The Fuzzy Gene Filter's success is ascribed to its incorporation of both parametric and non-parametric data features and its ability to be optimised to suit the specific data-set under analysis.