Research Article
Dimensionality Reduction Performance of Sparse PCA Methods
@INPROCEEDINGS{10.1007/978-3-030-92942-8_12, author={Thanh Do Van}, title={Dimensionality Reduction Performance of Sparse PCA Methods}, proceedings={Nature of Computation and Communication. 7th EAI International Conference, ICTCC 2021, Virtual Event, October 28--29, 2021, Proceedings}, proceedings_a={ICTCC}, year={2022}, month={1}, keywords={Dimensionality reduction PCA Sparse PCA}, doi={10.1007/978-3-030-92942-8_12} }- Thanh Do Van
Year: 2022
Dimensionality Reduction Performance of Sparse PCA Methods
ICTCC
Springer
DOI: 10.1007/978-3-030-92942-8_12
Abstract
Today, a hot topic is building forecasting models on large data sets of economic-financial time-series predictors using dimensionality reduction methods. The forecasting models built based on the dynamic factor model in which the factors extracted by the principal component analysis (PCA for short) method or sparse PCA (SPCA) method are superior to other benchmark models in terms of the forecast accuracy of models. Many pieces of literature have considered that the dimensionality reduction performance of the SPCA method seems to be higher than that of the PCA method. However, there have been no studies comparing the dimensionality reduction performance of those two methods to date.
The purpose of this article is to overcome that inadequacy by experimentally evaluating the dimensionality reduction performance of the two methods mentioned above on ten real-world data sets. Difference from previous beliefs, the experimental results show that the dimensionality reduction performance of the PCA and SPCA methods is competitive.
