| Autor: |
Akimov, V.A., Minkin, V.A. |
| Jazyk: |
angličtina |
| Rok vydání: |
2020 |
| Předmět: |
|
| DOI: |
10.25696/elsys.vc4.en.06 |
| Popis: |
Investigated various options for constructed artificial neural networks (ANNs) used to discriminate databases of behavioral parameters captured by vibraimage technology for patients with confirmed diagnosis of COVID-19 and reference group with confirmed absence of COVID-19 disease. The developed ANNs were learned using ADAM and Nesterov methods. The dependences of method accuracy and the number of errors (discriminating ability of the test) on the structure of ANN and the set of behavioral parameters are presented. Statistical analysis of the same groups of patients and the control group was carried out using standard statistical methods (mat expectation, SD, variability) and groups discrimination by ANN methods. The structure of ANN and input data of behavioral parameters was optimized. Achieved zero error of existing databases discriminating for the patients with a confirmed diagnosis of COVID-19 and the control group. Identified significant behavioral parameters for the diagnosis of COVID-19. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
