Support vector machines-kernel algorithms for the estimation of the water supply in Cyprus

Authors:
Fotis Maris;Lazaros Iliadis;Stavros Tachos;Athanasios Loukas;Iliana Spartali;Apostolos Vassileiou;Elias Pimenidis
Affiliations:
Democritus University of Thrace Greece;Democritus University of Thrace Greece;Aristotle University of Thessaloniki Greece;3University of Thessaly Greece;Aristotle University of Thessaloniki Greece;Democritus University of Thrace Greece;University of East London, UK
Venue:
ICANN'10 Proceedings of the 20th international conference on Artificial neural networks: Part II
Year:
2010

Citing 6
Cited 0

A training algorithm for optimal margin classifiers

COLT '92 Proceedings of the fifth annual workshop on Computational learning theory
Learning and Soft Computing: Support Vector Machines, Neural Networks, and Fuzzy Logic Models

Learning and Soft Computing: Support Vector Machines, Neural Networks, and Fuzzy Logic Models
An Artificial Neural Network model for mountainous water-resources management: The case of Cyprus mountainous watersheds

Environmental Modelling & Software
LIBSVM: A library for support vector machines

ACM Transactions on Intelligent Systems and Technology (TIST)
Fuzzy Weighted Support Vector Regression With a Fuzzy Partition

IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics
On the stability issues of linear Takagi-Sugeno fuzzy models

IEEE Transactions on Fuzzy Systems

Quantified Score

Hi-index	0.00

Visualization

Abstract

This research effort aimed in the estimation of the water supply for the case of "Germasogeia" mountainous watersheds in Cyprus. The actual target was the development of an ε-Regression Support Vector Machine (SVMR) system with five input parameters. The 5-Fold Cross Validation method was applied in order to produce a more representative training data set. The fuzzy-weighted SVR combined with a fuzzy partition approach was employed in order to enhance the quality of the results and to offer an optimization approach. The final models that were produced have proven to perform with an error of very low magnitude in the testing phase when first time seen data were used.