A novel fusion approach based on induced ordered weighted averaging operators for chemometric data analysis

摘要: This paper proposes a novel approach for the estimation of spectroscopic data by combining predictions an ensemble estimators using induced ordered weighted averaging (IOWA) fusion operators. For generation, we use Gaussian process regression (GPR) and extreme learning machine (ELM) associated with different kernels. To render model selection issue ELM as efficiently in GPR Bayesian method, develop automatic solution based on powerful differential evolution (DE) algorithm. During process, IOWA operator needs two things: (1) order-inducing value; (2) way to determine its weights. value, propose residual each estimated output value. Because cannot compute true residual, explore idea estimating residuals themselves associating estimator second same kind called estimator. learn weights these nonlinear operators, proposed method relies concept prioritized aggregation, where generate directly from residuals. Experimental results obtained three real datasets confirm interesting capabilities method. Copyright © 2013 John Wiley & Sons, Ltd.