Abstract

The authors of this study have built an outstanding data mining model for the classification of respiratory issues in children and adults, which they have applied in their research. Deep learning ensembles are built by utilising support vector regression (SVR), long short-term memory neural networks (LSTMs), and a metaheuristic optimization strategy that incorporates nonlinear learning in the DL ensemble (MHO). A collection of LSTMs with variable hidden layers and neurons is used to detect and exploit the underlying relationships in order to overcome the limitations of a single deep learning approach limited generalisation skills and robustness when faced with diverse input. The LSTM classification is then combined with a nonlinear-learning SVR and MHO to optimise the top-layer parameters. Nonlinear-learning meta-layer and LSTM classification. Finally, the final classification of the ensemble is provided by the fine-tuning meta-layer. Using data from six benchmark studies as well as energy consumption data sets, the proposed EDL is put to the test in two classification scenarios: ten-ahead and one-ahead classification.