ML-based group method of data handling: an improvement on the conventional GMDH

Abstract Machine learning (ML) has been recognized as a feasible and reliable technique for the modeling of multi-parametric datasets. In real applications, there are different relationships with various complexities between sets inputs their corresponding outputs. As result, models have developed levels complexity in input–output relationships. The group method data handling (GMDH) employs family inductive algorithms computer-based mathematical grounded on combination quadratic higher neurons certain number variable layers. this method, vector input features is mapped to expected response by creating multistage nonlinear pattern. Usually, each neuron GMDH considered partial function. paper, basic structure adapted changing functions enhance ability. To accomplish this, popular ML that shown reasonable function approximation performance, such support regression random forest, used, polynomial replaced these models. feasibility validity ML-based confirmed computer simulation.