Improved Sparrow Search Algorithm with the Extreme Learning Machine and Its Application for Prediction

The prediction accuracy and generalization ability of extreme learning machine (ELM) are reduced by randomly generated weight and threshold before training. To solve these problems, an improved sparrow algorithm (ISSA)-ELM hybrid prediction model is proposed in this paper. Firstly, SSA is improved by the opposition-based learning (OBL), cosine inertia weights and Levy flight strategy. Secondly, unimodal and multimodal benchmark functions are used to verify the performance of ISSA. The test results show that ISSA has better optimization accuracy, stability and statistical properties than SSA, whale optimization algorithm (WOA), particle swarm optimization algorithm (PSO), gravity search algorithm (GSA), grasshopper optimization algorithm (GOA) and dragonfly algorithm (DA). Finally, ISSA is applied to optimize the weights and thresholds of ELM, and four regression datasets are used to test ISSA-ELM. The simulation results show that ISSA-ELM has better prediction accuracy, generalization ability and stability than SSA-ELM and ELM.