Medical machine learning based on multiobjective evolutionary algorithm using learning decomposition

• Published in: Expert systems with applications Vol. 216; p. 119450
• Main Authors: Wang, Mingjing, Li, Xiaoping, Chen, Long, Chen, Huiling
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 15.04.2023
• Subjects: Cancer gene expression data sets; Evolutionary decomposition algorithm; Harris hawks learning; Lupus nephritis; Medical machine learning; Multi-objective; Pulmonary hypertension; Harris hawks learning; Evolutionary decomposition algorithm; Cancer gene expression data sets; Multi-objective; Lupus nephritis; Medical machine learning; Pulmonary hypertension
• ISSN: 0957-4174
• DOI: 10.1016/j.eswa.2022.119450

Medical machine learning technology has garnered great attention from both the computer and medical fields. In this study, a multi-objective evolutionary algorithm integrating decomposition and harris hawks learning (MOEA/D-HHL) is presented for medical machine learning; harris hawks learning can guarantee a good variety and systematic MOEA/D-HHL solutions. The performance of this MOEA/D-HHL is first evaluated using these benchmarks (DTLZ1-DTLZ7). In addition, MOEA/D-HHL is used to construct machine learning algorithms for medical cancer gene expression data sets with the following three objectives in mind: selected features, classification accuracy, and correlation measures. The MOEA/D-HHL is finally applied efficiently to the clinical data of lupus nephritis and pulmonary hypertension with the best NMI of 0.9652 and ARI of 0.9749 values on lupus nephritis, and the best NMI of 0.9686 and ARI of 0.9742 values on pulmonary hypertension respectively. On clinically relevant data for lupus nephritis and pulmonary hypertension, the experimental results indicate that the proposed MOEA/D-HHL algorithm outperforms current methods. The statistical results demonstrate that all metrics have predictive capabilities and that the suggested MOEA/D-HHL is more stable for an emerging medical machine learning framework. MOEA/D-HHL may be seen as a promising computer-assisted approach for medical machine learning development. •The MOEA/D-HHL is presented for medical ML.•Performance of the MOEA/D is boosted by HHL.•The better balance between intensification and diversification has been guaranteed.•The MOEA/D-HHL is verified on DTLZ1-DTLZ7 benchmark.•MOEA/D-HHL may be used for medical machine learning successfully.