Journal
| Title | On the Discrete Choice Model with High-Dimensional Data using Orthogonal Sparse PCA and Super Sparse PCA Posted by Bernadette Tubo |
| Authors | Farlley G. Bondaug, Bernadette F. Tubo, Oliver R. Capilitan and Calixto G. Elnas, Jr. |
| Publication date | 2021 |
| Journal | Asia-Pacific Journal of Science, Mathematics and Engineering (APJSME) |
| Volume | 7 |
| Issue | 2 |
| Pages | 19-24 |
| Publisher | Office of the Vice Chancellor for Research and Enterprise, Mindanao State University-Iligan Institute of Technology (MSU-IIT) |
| Abstract | The study introduces a two-step procedure on dimension reduction and function approximation for discrete choice model with high-dimensional predictors. The procedure is a synthesis of two techniques, namely: Super Sparse Principal Components Analysis (SSPCA) and Generalized Additive Model (GAM). The methodology is an extension of the work of Lee, et. al (2009) on SSPCA where the results are further processed in a classification method with GAM labeled in this study as SS-GAM. Likewise, Orthogonal Sparse PCA with GAM is also considered (OS-GAM). Moreover, the General Additive Sparse (GAS) with GAM (GAS-GAM) is used as a baseline procedure reference. The simulation study reveals that in terms of predictive ability and model size, SS-GAM performed better compared to OS-GAM and GAS-GAM. However, in terms of computational time, OS-GAM seems to be not affected by the increase in dimension in the features. Also, for data with low correlation, OS-GAM gives the sparsest model. |
| Index terms / Keywords | principal components analysis, orthogonal sparse PCA, super sparse PCA, general additive sparse, generalized additive model, discrete choice model |
| URL | ISSN: 2244-5471 |