Aggregation of randomized model ensemble outcomes for reconstructing nuclear signals from faulty sensors

In the approach here adopted, the signals groups are randomly generated resorting to the Random Feature Selection Ensemble (RFSE) technique (Bryll et al. 2003, Baraldi et al. 2009, Polikar 2006) and a corresponding number of Principal Component Analysis (PCA) regression models (Jolliffe 1986, Diamantaras & Kung 1996, Scholkopf 1999, Moore 1981) are developed based on the signals of the individual groups. The randomized group generation procedure allows obtaining highly diverse signals groups and, correspondingly, diverse signal predictions as outcomes from the individual models.