Kernel logistic PLS: A tool for supervised nonlinear dimensionality reduction and binary classification

• DOI: 10.1016/j.csda.2007.01.004
• Publication Date: 2007-01-17T12:20:12Z
• Authors: Arthur Tenenhaus; Alain Giron; Emmanuel Viennet; Michel Béra; Gilbert Saporta; Bernard Fertil
• Keywords: Kernel (algebra); Binary classification; Benchmark (surveying)

''Kernel logistic PLS'' (KL-PLS) is a new tool for supervised nonlinear dimensionality reduction and binary classification. The principles of KL-PLS are based on both PLS latent variables construction and learning with kernels. The KL-PLS algorithm can be seen as a supervised dimensionality reduction (complexity control step) followed by a classification based on logistic regression. The algorithm is applied to 11 benchmark data sets for binary classification and to three medical problems. In all cases, KL-PLS proved its competitiveness with other state-of-the-art classification methods such as support vector machines. Moreover, due to successions of regressions and logistic regressions carried out on only a small number of uncorrelated variables, KL-PLS allows handling high-dimensional data. The proposed approach is simple and easy to implement. It provides an efficient complexity control by dimensionality reduction and allows the visual inspection of data segmentation.