Improving the Robustness of Prediction Model by Transfer Learning for Interference Suppression of Electronic Nose

This paper gives a solution to solve the interference problem of electronic nose (e-nose), which is ill-posed due to the uncertainty and unpredictability of its instable behavior. Traditional methods for interference suppression are component correction frameworks, which are laborious or little efficient. With interference (especially background interference and sensor drift), the distribution of test data obtained in practical application is different from that of the previous training data. From the viewpoint of machine learning, a novel domain correction and adaptive extreme learning machines (DC-AELM) framework with transferring capability is proposed to solve the serious interference problem in e-nose. The framework consists of two parts: 1) DC, which makes the distributions of two domains close and 2) AELM, which realizes the knowledge transfer at the decision level and makes the robustness of the prediction model improved. This method is motivated by the idea of transfer learning, especially from the perspective of domain correction and decision-making, to realize the knowledge transfer for interference suppression. A background interference data set obtained by our designed e-nose and a public benchmark sensor drift data set are used to verify the effectiveness of the proposed DC-AELM method.