Improved extreme learning machine based on adaptive dual-strategy optimization algorithm and its application

Abstract Aiming at the problem of low accuracy and unstable prediction results caused by the random connection weight threshold between the input layer and the hidden layer of extreme learning machine (ELM). An adaptive dual-strategy improved pelican optimization algorithm (IPOA) -ELM regression prediction model is proposed. Firstly, the pelican optimization algorithm (POA) is improved by the Logistic-Tent chaotic map, improved convergence factor by adaptive double strategy method and reverse learning strategy. Then, the performance of IPOA is verified by multiple groups of multi-dimensional single-peak and multi-peak test functions. The test results show that IPOA has better accuracy, stability, and robustness than POA, butterfly optimization algorithm (BOA), cuckoo algorithm (CS), grey wolf algorithm (GWO), particle swarm optimization algorithm (PSO), genetic algorithm (GA) and mouse swarm optimization algorithm (RSO). Finally, IPOA is applied to optimize ELM, and the performance of the IPOA-ELM model is verified by three engineering data sets. The simulation results show that the convergence accuracy, stability, and robustness of the IPOA-ELM model are better than those of the POA-ELM and ELM models.