DAILY PAPER REVIEW

180706_Support Vector Machines and Multilayer Perceptron Networks Used to Evaluate...

1. Title, Journal and Authors

- Title: Support Vector Machines and Multilayer Perceptron Networks Used to Evaluate the Cyanotoxins Presence from Experimental Cyanobacteria Concentrations in the Trasona Reservoir (Northern Spain)

- Journal : Water Resour Manage

- Authors : J. A. Vilán Vilán1, J. R. Alonso Fernández2, P. J. García Nieto3, F. Sánchez Lasheras4, F. J. de Cos Juez5, C. Díaz Muñiz2

- Affiliation :

1: Department of Mechanical Engineering, University of Vigo, 36200 Vigo, Spain

2: Cantabrian Basin Authority, Spanish Ministry of Agriculture, Food and Environment, 33071 Oviedo, Spain

3: Department of Mathematics, Faculty of Sciences, University of Oviedo, 33007 Oviedo, Spain

4: Department of Construction and Manufacturing Engineering, University of Oviedo, 33204 Gijón, Spain

5: Mining Exploitation and Prospecting Department, University of Oviedo, 33004 Oviedo, Spain

 

2. Summary

- This paper suggested a methodology to predict concentration of cyanotoxins using support vector machines (SVM) and multilayer perceptron networks (MLP).

- The data used for the SVM and MLP consisted of 8 biological variables and 15 physical-chemical input variables.

- Cross validation was used for finding a suitable set of hyper parameters of SVM and MLP.

- The SVM with the radial basis kernel function (RBF-Kernel) is showed the best performance for predicting the concentration of cyanotoxins in the Trasona reservoir.

- Two dominant species of the cyanobacteria had synergistic behavior on the increase of cyanotoxins concentration.

 

3. Application to research

- The cross validation is the standard technique to find a suitable set of hyper parameters of machine learning.

- The significance of each biological and physical-chemical variables on the concentration of cyanotoxins is presented through the results of this paper.

- The sensitivity analysis is essential to optimize the input variables of machine learning and to provide a scientific basis for selecting variables.

 

4. Contact

Heewon Jeong (Ph.D. program)

Environmental Systems Engineering Lab.

School of Environmental Science & Engineering

Gwangju Institute of Science and Technology

1 Oryong-dong Buk-gu Gwangju, 500-712, Korea

Email :  gua01114@gist.ac.kr

첨부 (0)