Neural Network Ensembles for Online Gas Concentration Estimation Using an Electronic Nose
Ensemble method is a learning paradigm that has been shown to improve the performance of classical learning methods which are based on single model. However, for an ensemble method to be effective, it is essential that the base models are sufficiently accurate and error-independent (i.e. diverse) in their predictions. Moreover, ensemble integration is one of the most critical steps in ensemble learning. In this paper, a dynamic integration method is proposed and applied in electronic nose for online concentration estimation of some indoor air pollutants namely formaldehyde, benzene, toluene, and carbon monoxide. For comparison purpose, other integration methods were also evaluated. Experimental results show that this method is attractive, and with additional improvement it can be a good alternative for online air quality monitoring using electronic nose systems.
Keywords: electronic nose, neural network ensembles, dynamic integration method, online monitoring
Download Full-Text
ABOUT THE AUTHORS
Chaibou Kadri
College of Communication Engineering, Chongqing University
Fengchun Tian
College of Communication Engineering, Chongqing University
Lei Zhang
College of Communication Engineering, Chongqing University
Guojun Li
Chongqing Communication Institute
Lijun Dang
College of Communication Engineering, Chongqing University
Guorui Li
College of Communication Engineering, Chongqing University
Chaibou Kadri
College of Communication Engineering, Chongqing University
Fengchun Tian
College of Communication Engineering, Chongqing University
Lei Zhang
College of Communication Engineering, Chongqing University
Guojun Li
Chongqing Communication Institute
Lijun Dang
College of Communication Engineering, Chongqing University
Guorui Li
College of Communication Engineering, Chongqing University