A GA-Based Approach to ICA Feature Selection: An Efficient Method to Classify Microarray Datasets

Although many independent component analysis (ICA) based algorithms were proposed to tackle the classification problem of microarray data, a problem is usually ignored that which and how many independent components can be used to best describe the property of the microarray data. In this paper, we proposed a GA approach for IC feature selection to increase the classification accuracy of two different ICA based models: penalized independent component regression (P-ICR) and ICA based Support Vector Machine (SVM). The corresponding experimental results are listed to show that the IC selection method can further improve the classification accuracy of the ICA based algorithms.