A global method for identifying dependences between helio-geophysical and biological series by filtering the precedents (outliers)
| العنوان: | A global method for identifying dependences between helio-geophysical and biological series by filtering the precedents (outliers) |
|---|---|
| المؤلفون: | Yu. I. Gurfinkel, T. A. Matveeva, T. K. Breus, V. A. Ozheredov |
| المصدر: | Izvestiya, Atmospheric and Oceanic Physics. 50:793-804 |
| بيانات النشر: | Pleiades Publishing Ltd, 2014. |
| سنة النشر: | 2014 |
| مصطلحات موضوعية: | Gray code, Atmospheric Science, Mathematical optimization, Polynomial, Optimization problem, Series (mathematics), Outlier, Monte Carlo method, Genetic algorithm, Sorting, Oceanography, Algorithm, Mathematics |
| الوصف: | A new approach to finding the dependence between heliophysical and meteorological factors and physiological parameters is considered that is based on the preliminary filtering of precedents (outliers). The sought-after dependence is masked by extraneous influences which cannot be taken into account. Therefore, the typically calculated correlation between the external-influence (x) and physiology (y) parameters is extremely low and does not allow their interdependence to be conclusively proved. A robust method for removing the precedents (outliers) from the database is proposed that is based on the intelligent sorting of the polynomial curves of possible dependences y(x), followed by filtering out the precedents which are far away from y(x) and optimizing the coefficient of nonlinear correlation between the regular, i.e., remaining, precedents. This optimization problem is shown to be a search for a maximum in the absence of the concept of gradient and requires the use of a genetic algorithm based on the Gray code. The relationships between the various medical and biological parameters and characteristics of the space and terrestrial weather are obtained and verified using the cross-validation method. It is proven that, by filtering out no more than 20% of precedents, it is possible to obtain a nonlinear correlation coefficient of no less than 0.5. A juxtaposition of the proposed method for filtering precedents (outliers) and the least-square method (LSM) for determining the optimal polynomial using multiple independent tests (Monte Carlo method) of models, which are as close as possible to real dependences, has shown that the LSM determination loses much in comparison to the proposed method. |
| تدمد: | 1555-628X 0001-4338 |
| DOI: | 10.1134/s0001433814080064 |
| URL الوصول: | https://explore.openaire.eu/search/publication?articleId=doi_________::295e13ca75bf37d354c2566d40a4a123 https://doi.org/10.1134/s0001433814080064 |
| Rights: | CLOSED |
| رقم الانضمام: | edsair.doi...........295e13ca75bf37d354c2566d40a4a123 |
| قاعدة البيانات: | OpenAIRE |
| تدمد: | 1555628X 00014338 |
|---|---|
| DOI: | 10.1134/s0001433814080064 |