Feature extraction plays a fundamental role in the KDD and Data Mining process. There are many algorithms for mining data based on Principal Component Analysis (PCA), a powerful statistical tool which is identical to the Karhunen-Loeve transform for pattern recognition. Independent Component Analysis (ICA) is a recently developed technique based on the assumption of statistical independence between the components that acts as a remedy to the limitations of PCA. In this paper, some applications of ICA in the KDD process and in the Data Mining step of this process are described. It is proposed a fuzzy method to quantify the information from a linear combination of input data and a genetic algorithm to find the components with the optimal values of such measure.
展开▼
