Drift compensation for E-nose using QPSO-based domain adaptation kernel ELM

Abstract

A novel theoretical framework for drift compensation and classification of an electronic nose (E-nose), called QPSO-based domain adaptation kernel extreme learning machine (QDA-KELM) is presented in the work. The kernel method combines with domain adaption extreme learning machine (DAELM) to remove the drift in E-nose and enhance the classification performance. A swarm intelligent algorithm is utilized for the optimization of the model parameters. In order to evaluate the performance of our approach, three types of common kernels are used to form the composite kernel function. In addition, ELM and DAELM are compared with the proposed method. Finally, we also applied Analysis of Variance (ANOVA) to demonstrate our results are significantly better than the control methods.