Evolving an optimal kernel extreme learning machine by using an enhanced grey wolf optimization strategy

Since its introduction, kernel extreme learning machine (KELM) has been widely used in a number of areas. The parameters in the model have an important influence on the performance of KELM. Therefore, model parameters must be properly adjusted before they can be put into practical use. This study proposes a new parameter learning strategy based on an improved grey wolf optimization (IGWO) strategy, in which a new hierarchical mechanism was established to improve the stochastic behavior, and exploration capability of grey wolves. In the proposed mechanism, random local search around the optimal grey wolf was introduced in Beta grey wolves, and random global search was introduced in Omega grey wolves. The effectiveness of IGWO strategy is first validated on 10 commonly used benchmark functions. Results have shown that the proposed IGWO can find good balance between exploration and exploitation. In addition, when IGWO is applied to solve the parameter adjustment problem of KELM model, it also provides better performance than other seven meta-heuristic algorithms in three practical applications, including students' second major selection, thyroid cancer diagnosis and financial stress prediction. Therefore, the method proposed in this paper can serve as a good candidate tool for tuning the parameters of KELM, thus enabling the KELM model to achieve more promising results in practical applications. (C) 2019 Elsevier Ltd. All rights reserved.