A Hybrid Fuzzy-SVM classifier, applied to gene expression profiling for automated leukaemia diagnosis

A hybrid fuzzy-SVM classifier, used to automate leukaemia diagnosis based on microarray gene expression data, is presented. A publicly available dataset was used to develop and test the classifier. A fuzzy gene filter was developed to select the genes which show significant class variation between various leukaemia types. The results obtained from using all the genes for classification is compared to those obtained when only the top 25 differentiating genes are used. The filtered gene classifier was able to correctly classify the entire test dataset, compared to the unfiltered gene classifier which was only able to achieve an accuracy of 84.2%. The results show that, by reducing dimensionality, classification accuracy is improved since redundant information is excluded, thereby limiting the effect of potential outliers.