An improvised nature-inspired algorithm enfolded broad learning system for disease classification

Abstract

Deep analysis of genomic data reveals that many deadly diseases are generated due to genetic mutation. To make the health care system more robust, a machine learning researcher’s prime intention is to classify the genomic data more efficiently within less time. As the genomic data suffers from the malediction of excessive dimensionality, the selection of the notable genes is always a big challenge for the researcher. The selection of prominent genomic key attributes by any nature-inspired learning algorithm always remains a non-deterministic polynomial-time (NP-Hard) problem. Therefore, there is always a scope to apply new algorithms. In this projected work, an improvised sine–cosine hybridized Monarch Butterfly Optimization (SC-MBO) algorithm, is embedded with the Broad Learning System (BLS), which is defined as SC-MBO-BLS, for choosing the most significant genes and classifying the genomic data simultaneously. Initially, Kernel-based Fisher Score (K-FS) is applied to select notable genes. Then, the selected genes further undergoes for execution using the SC-MBO-BLS model. To prove the effectiveness of the suggested model, ten cancerous genomic data are considered. Here, several performance evaluators (i.e., precision, MCC, sensitivity, Kappa, F-score, and specificity) are applied for unbiased comparison. This presented model is compared with SC-MBO wrapped Multilayer Perceptron (SC-MBO-MLP), SC-MBO wrapped Extreme Learning Machine (SC-MBO-ELM), and SC-MBO wrapped Kernel Extreme Learning Machine (SC-MBO- KELM) and yields the highest accuracy in ten datasets such as 100%, 98.4%, 99%, 99.6%, 100, 97.2%, 100%, 100%, 98.6%, 99.5% in Leukemia, Colon tumor, Breast cancer, Ovarian cancer, Lymphoma-3, MLL, ALL-AML-3, SRBCT, ALL-AML-4 and Lung cancer respectively. Further, the existing twenty standard models are taken for comparison with the suggested model. Additionally, to assess the presented model, a statistical method i.e., Analysis of variance (ANOVA) is considered. As per the above quantitative and qualitative estimation, it is deduced that the suggested SC-MBO-BLS approach outclasses other considering models.

Keywords

• Genomic data
• High dimensionality
• Notable genes
• Feature extraction
• Kernel fisher score
• Classification