Nonlinear adaptive regression predictors based on singular decomposition |
| |
| Authors: | Jan John Josef ?tekl |
| |
| Institution: | (1) Faculty of Electrical Engineering, Department of Automatic Control, The Czech Technical University, ČSFR;(2) Institute of the Physics of the Atmosphere, Czechosl. Acad. Sci., Prague |
| |
| Abstract: | Summary The paper deals with a non-linear regression model, linear in parameters. The least-squares adaptation method has been used
to determine the model parameters. The conditionality problem in solving equations, which follows from the dependent predictors,
has been suppressed by singular decomposition. This model is especially suitable for treating meteorological problems because
non-linear processes, e.g. changes of air masses, changes of circulation patterns, etc., can be treated. The model has been
tested in connection with the forecast of daily maximum and minimum temperatures. The RMS error lies in the range of 1–2°C,
if the principal PPM is supposed.
Резюме Оnuсывеamся мamемamuческuŭ annaрam нелuнеŭноŭ ре?rt;рессuонноŭ мо?rt;елu, лuнеŭноŭ оmнос umельно naрaмеmров. Для оnре?rt;еленuя
naрaмеmров мо?rt;елu uсnользуеmся меmо?rt; нauменьщuх квa?rt;рamов. Плохaя обусловленносmь сuсmемы нauменьщuх квa?rt;рamов,
коmорaя вызвaнa взauмноŭ зaвuсuмосmью nре?rt;uкmоров, всmрaняеmся nрu nомощu сuн?rt;улярноŭ ?rt;екомnозuцuu. Эma мо?rt;ель
nо?rt;хо?rt;um ?rt;ля nрuмененuя в меmеороло?rt;uu, maк кaк онa nозволяеm рaбоmamь с нелuнеŭносmямu, кaк нanрuмер uзмененuе
воз?rt;ущных мaсс, uзмененuе хaрaкmерa цuркуляцuu u m. n. Мо?rt;ель nроверяеmся нa nро?rt;нозе ?rt;невных мuнuмaльных u мaксuмaльных
mемnерamур. Сре?rt;няя квa?rt;рamuческaя ошuбкa нaхо?rt;umся в uнmе?rt;рaле оm 1 ?rt;о 2° (uсnользовaн nрuнцun РРМ).
|
| |
| Keywords: | |
| 本文献已被 SpringerLink 等数据库收录！ |
|
