A Hybrid Encoded Based Extreme Learning Machine Based On Particle Swarm Optimization Algorithm With Salp Swarm Algorithm

P. SrinivasaRao, G. Apparao Naidu, Nageswara Rao, K. Ramakrishna

The classical ELM training algorithm is a non-skilled and random process that could not converge efficiently with high performance or the potential imprisonment of the local minima problem. The ELM model for the monthly river flow is integrated with a newly expanded meta-heuristic algorithm (e.g. the Salp Swarm Algorithm (SE)). As a case study is used twenty years in the Tigris river data sequence at Baghdad Station in Iraq. For the construction of the predictive models based on the antecedent values, different input combinations are used. The results are assessed on the basis of various statistical measurements and graphical presentations. The SSA-ELM river flow forecasts have surpassed the classic ELM and other AI models. During the testing process, a sufficient improvement in the level accuracies of the proposed SSA-ELM model was made compared to the classic ELM model (8.4 and 13.1 percent increase in RMse and MAE respectively).

Volume 12 | Issue 6

Pages: 283-289

DOI: 10.5373/JARDCS/V12I6/S20201031