Title | ||
---|---|---|
Sur l'utilisation conjointe de la régression sur co mposantes principales et des ondelettes |
Abstract | ||
---|---|---|
The principal components regression (PCR) is an applied regression on PCA factors of a PCA beforehand carried out on initially strongly correl ated variables. The use of the PCA allows replacing the initial variables, by principle compo nents which preserve the quasi-total of information, and which present the advantage to be non-correlated. These components are taken as explanatory variables for a multiple linea r regression. The PCR modeling quality remains affected by the existence of noise in the i nitial variables. In this work, we propose a denoising of the data by wavelets (thresholding) ma king possible separation of the signal from the noise without losing information. We show usin g French stock-exchange data, that the elimination of the noise on the main constituents b y a soft thresholding based on wavelets improves the quality of adjustment of the regressio n model as well as forecast quality. We confirm result by simulation. |
Year | Keywords | DocType |
---|---|---|
2010 | denoising,pca,pcr,simulation,: wavelets,thresholding | Journal |
Volume | Citations | PageRank |
41 | 0 | 0.34 |
References | Authors | |
1 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Salwa BENAMMOU | 1 | 0 | 0.34 |
Nabiha HAOUAS | 2 | 0 | 0.34 |
Zied KACEM | 3 | 0 | 0.34 |