Title: Incorporating sparsity, smoothness and group structure in regularized models for spectroscopic data
Authors: Ying Zhu - National Institute of Education, Nanyang Technological University (Singapore)
Chin Gi Soh - Nanyang Technological University (National Institute of Education) (Singapore) [presenting]
Abstract: High-dimensional spectroscopic data has applications in many fields such as food science, forensic science and biomedical science due to the information it provides about the chemical compositions of the samples. The fitting of classification and regression models to such data is known to be a challenging task due to the high-dimensional setting, as well as the issue of high correlation between spectral variables. One method that has gained interest in recent years is the use of regularization to overcome these challenges. A regularized model for spectroscopic data is presented that incorporates sparsity, smoothness and group structure. The results from some simulation studies on the use of the regularized model will be discussed. An application of the model to Fourier-transform infrared spectroscopic data adulteration studies in olive oil will also be presented. The results suggest that the sparse fused group lasso is able to achieve good prediction performance, while improving on the interpretability of the resulting models.