A problem of Significant Wave Height prediction is tackled.
A hybrid GGA-ELM algorithm is proposed.
The GGA-ELM approach searches for the best set of features in the problem.
Experiments in real data of buoys located at the Western coast of the USA are carried out.
This paper proposes a novel hybrid approach for feature selection in two different relevant problems for marine energy applications: significant wave height (Hm0) and wave energy flux (P ) prediction. Specifically, a hybrid Grouping Genetic Algorithm – Extreme Learning Machine approach (GGA-ELM) is proposed, in such a way that the GGA searches for several subsets of features, and the ELM provides the fitness of the algorithm, by means of its accuracy on Hm0 or P prediction. Since the GGA was specifically created for problems involving a number of groups, the proposed algorithm may be used to evolve different groups of features in parallel, which may improve the performance of the predictions obtained. After the feature selection process with the GGA-ELM, the final results are given by an ELM and also by a Support Vector Machine, both working on the best GGA groups obtained. The performance of the proposed system has been tested in a real problem of Hm0 and P prediction at the Western coast of the USA, obtaining good results.
- Wave energy flux;
- Marine energy;
- Significant wave height;
- Grouping genetic algorithm (GGA);
- Extreme Learning Machines;
- Support vector machines
© 2016 Elsevier Ltd. All rights reserved.