Title :
An ANN - PCA adaptive forecasting model
Author :
Nastac, Dumitru Iulian ; Cristea, Paul Dan
Author_Institution :
Electron., Dept., Univ. Politeh. of Bucharest, Bucharest, Romania
Abstract :
The paper describes specific aspects that concern Principal Component Analysis (PCA) when using it as a preprocessing tool in a forecasting model. Principal component analysis is an efficient used statistical technique for dimensional reduction, and here we employ the PCA to decorrelate the input data before training a neural network architecture. This approach reveals important regularities in the PCA transformation matrix that can improve the entire model.
Keywords :
data reduction; forecasting theory; learning (artificial intelligence); neural nets; principal component analysis; ANN-PCA adaptive forecasting model; PCA transformation matrix; dimensional reduction; forecasting model; neural network architecture training; preprocessing tool; principal component analysis; statistical technique; Adaptation models; Bioinformatics; Forecasting; Genomics; Principal component analysis; Time series analysis; Training;
Conference_Titel :
Systems, Signals and Image Processing (IWSSIP), 2012 19th International Conference on
Conference_Location :
Vienna
Print_ISBN :
978-1-4577-2191-5
