Dimension reduction for a multivariate time series process of a regenerative glass furnace
- Autores: Xuan Huang, Davit Khachatryan
- Localización: Journal of quality technology: A quarterly journal of methods applications and related topics, ISSN 0022-4065, Nº. 50, 1, 2018, págs. 98-116
- Idioma: inglés
- Enlaces
-
Resumen
Conventional dimension-reduction methods for multivariate time series have been based on the inherent assumption that the noise is white and therefore uninformative. While this assumption may have been realistic for a number of applications, it is violated in the case of a large industrial furnace that we discuss in this article. The article introduces a novel method that is used for dimension reduction of a multivariate time series process for which the underlying common factors as well as the contaminating noise are autocorrelated. Using simulation studies, it is shown that the proposed method uncovers the common factors more accurately than some of the other conventional methods. Applying the method to the furnace data also correctly reveals features of the process that other methods fail to capture. An important feature of the proposed method is that it is practical and requires neither extensive statistical modeling nor deep subject-matter knowledge of the data-generating process.
