A new regression method based on independent component analysis

Blind decomposition of infrared spectra using flexible component analysis

“…Alternatives to library matching approach are blind decomposition methods, wherein pure components' spectra are extracted using mixtures spectra only. Blind approaches to pure components spectra extraction have been reported in NMR spectroscopy [1], infrared (IR) [2][3][4] and near infrared (NIR) spectroscopy [4][5][6], EPR spectroscopy [7,8], mass spectrometry [4, 9,1 0] Raman spectroscopy [11,12] etc. In a majority of blind decomposition schemes independent component analysis (ICA) [13][14][15] is employed to solve related blind source separation (BSS) problem.…”

“…The two requirements: to have more linearly independent mixtures than pure components and to have statistically independent pure components seem to be most critical for the success of the BSS approach to blind decomposition of the mixtures spectra into pure components spectra [4,5,8,10]. Statistical independence assumption is certainly not fulfilled in the case of IR spectra [2][3][4][5][6] because they are highly correlated i.e. overlapped.…”

“…overlapped. Raw data preprocessing technique by first or second order derivative has been used in FT-IR spectra analysis to reduce level of statistical dependence among pure components, [2][3][4][5][6]. This technique actually belongs to the generalization of the ICA known as dependent component analysis (DCA), [14,[16][17][18].…”

Prediction Model of Air Pollutant Levels Using Linear Model with Component Analysis

“…ICA establishes independent components from original variables. The concept of ICA is regarded to be able in explaining more for variable relationship because independence is a high-order statistic that is in favor over orthogonality [9]. The GLS regression forms relationship between response variable (y) and the ICs from ICA along with other explanatory variables (e.g., days in week, season).…”

“…Even though these two methods have their own approach, the goal is similar is to build components that are statistically independent with each other. In regression analysis, this is particularly very useful and become good input as predictors in a regression model since they optimize spatial patterns and remove complexity due to multicollinearity [8], [9]. ICR and PCR have been widely used in particular for plant study [9], dam deformation study [10], air pollutants in subway [11], air quality management [2], [3], [12], and O 3 prediction [1].…”