Optimised feature selection and entropy-based graph classification of gene expression data
| Autor: | Audu Musa Mabu, Raghav Yadav, Rajesh Prasad |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
Computer science
business.industry Data classification Biomedical Engineering Medicine (miscellaneous) Pattern recognition Feature selection Health Informatics Krill herd Euclidean distance Biomaterials ComputingMethodologies_PATTERNRECOGNITION Discriminative model Graph classification Outlier Entropy (information theory) Artificial intelligence business |
| Zdroj: | International Journal of Medical Engineering and Informatics. 12:357 |
| ISSN: | 1755-0661 1755-0653 |
| DOI: | 10.1504/ijmei.2020.108239 |
| Popis: | Gene expression (GE) profiles expansively revised to disclose intuition into the multifariousness of cancer furthermore to discover concealed information which provides biological knowledge for the classification of cancer. Precise cancer classification straightly through original GE profiles stays challenging on account of the intrinsic high-dimension feature along with the small magnitude of the data samples. Therefore, choosing high discriminative genes as of the GE data have turn into progressively fascinating in the bioinformatics field. This given paper gives a technique for the GE data classification utilising entropy-based graph classifier. Initially, the proposed technique evaluate the GE data's signal to noise ratio (SNR) values, additionally, selects the relevant features using krill herd (KH) optimization process. The truth is that not all features are helpful for classification, and some redundant together with the irrelevant features might even serve as outlier. To dispose the outliers, feature reduction is done with the assist of Euclidean distance. Classification is made utilising entropy-based graph classifier. The proposed process' effectiveness contrasted with the existing method concerning classifications is established from the experimental outcome. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|