Three-dimensional fluorescence spectra model optimisation for water quality analysis based on particle swarm optimisation

Authors:
Xiao-Li Wu;Jin-Rong Li;Jing Jie;Hui Zheng
Affiliations:
School of Automation and Electrical Engineering, Zhejiang University of Science and Technology, Hangzhou 310023, China;School of Automation and Electrical Engineering, Zhejiang University of Science and Technology, Hangzhou 310023, China;School of Automation and Electrical Engineering, Zhejiang University of Science and Technology, Hangzhou 310023, China;School of Automation and Electrical Engineering, Zhejiang University of Science and Technology, Hangzhou 310023, China
Venue:
International Journal of Wireless and Mobile Computing
Year:
2014

Citing 6
Cited 0

Original Contribution: Stacked generalization

Neural Networks
Stacked regressions

Machine Learning
An introduction to variable and feature selection

The Journal of Machine Learning Research
An improved multidimensional scaling localisation algorithm

International Journal of Wireless and Mobile Computing
Using group-decided Watts-Strogatz particle swarm optimisation to direct orbits of chaotic systems

International Journal of Wireless and Mobile Computing
Dynamic packet fragmentation based on particle swarm optimised prediction

International Journal of Wireless and Mobile Computing

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this paper, a model combination method is proposed to improve the model precision of water quality analysis with three-dimensional 3D fluorescence spectra. The key to successful model combination is the selection of sub-models, which also means selection of excitation wavelength for 3D fluorescence instrument miniaturisation. A particle swarm optimisation PSO algorithm is designed to select effective sub-models, in which the combinational model is built. Field samples from surface water and urban wastewater are used as research objects. Following the proposed PSO method, three excitation wavelengths were selected, and the corresponding sub-models were linearly combined to an optimised combinational model. The experimental results showed that the root mean square errors of prediction of the combinational model decreased significantly, whether compared with the sub-models having the best prediction precision or the combinational models without sub-models selection.