Data Mining with Independent Component Analysis
Fasong Wang, Hongwei Li, Rui Li · 2006
Independent component analysis (ICA)/blind source separation (BSS) has received many attentions in neural network and signal processing area recent years. In this paper, we consider the data mining problem with ICA. The data model of under-complete ICA in data mining is given and then gives the most popular ICA algorithm-natural gradient algorithm (NGA). Several applications of data mining with ICA is considered, such as latent variable decompositions, multivariate time series analysis and prediction, text document data analysis, extracting hidden signals in satellite images, weather data mining and so on. All these discussions suggest the huge potential outlook of data mining using ICA. The other contribution of this paper is it contains several literature surveys on various aspects of data mining using ICA