Enhancing extreme learning machines classification with moth-flame optimization technique

Abstract

Extreme Learning Machine (ELM) algorithm assigns the input weights and biases in a “one-time stamp” fashion, this method makes the algorithm to be ill-conditioned and reduces its classification accuracy. The contribution of this work is the enhancement of the performance of ELM with the Moth-Flame Optimization (MFO) algorithm to improve classification accuracy. A hybrid of the Moth-Flame Optimization and Extreme Learning Machine (MFO-ELM) algorithm is implemented in MATLAB. MFO ensures a concurrent simulation of exploration and exploitation of the search space to select an optimum candidate solution. The candidate solution is reshaped into input weights and biases for ELM classification. The hybrid algorithm is validated on five life-selected datasets. The performance improvement of MFO-ELM is compared with ELM-optimized Particle Swarm Optimization (PSO-ELM) and Competitive Swarm Optimization (CSO-ELM) algorithms. The improvement rates are qualitatively and quantitatively evaluated to show the improvement of MFO-ELM on ELM and the other meta-heuristic algorithms. MFO-ELM improved the accuracies of the basic ELM in all 100% of the simulations and performed better than the other meta-heuristic algorithms in 80% of the simulations. The performance of MFO-ELM is more competitive, and it is recommended for solving classification problems.

Analysis of new differential evolution variants to solve multi-modal problems

“…At the end, the performance of levels and subgroups are evaluated and non-performing levels and groups are merged into nearby levels or subgroups. The DESBS and HFCDE are run for twenty (20) times on all twenty-five (25) problems out of which five (5) problems are uni-modal and twenty (20) are multi-modal. All problems are minimization problems, hence the minimal value i.e.…”

“…The hybrid of the moth-flame optimization and extreme learning machine (MFO-ELM) algorithm [20] enhances the performance of traditional extreme learning machine (ELM) by a hundred percentage (100%). MFO-ELM is a classification algorithm and it 80 % more efficient than the compared algorithms.…”