Máté, Kormos (2016) Scarlatina: a time series analysis. Outstanding Student Paper, BCE, Statisztika és ökonometriai szekció. Szabadon elérhető változat / Unrestricted version: http://publikaciok.lib.uni-corvinus.hu/publikus/tdk/20160322145723.pdf
|
PDF
- Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
1MB |
Free and unrestricted access: http://publikaciok.lib.uni-corvinus.hu/publikus/tdk/20160322145723.pdf
Abstract
In this paper, some methods are taken into account to model the time series data of scarlet fever cases (also known as, by its latin name, scarlatina) in Hungary, relying on weekly data from 23rd March 1998 to 5th October 2015. The purpose of this paper is to select the most appropriate method for this task. The methods involved are machine learning techniques (boosting trees and support vector machine regression) and 'traditional' ones (nonlinear least squares, nonlinear Poisson, nonlinear negative binomial regression). First, discussed are the methods to be followed by the description of the exact models applied to the given data. Last, the comparison of the models is carried out on an independent dataset only to let the nonlinear Poisson regression emerge as the best one.
| Item Type: | Outstanding Student Paper |
|---|---|
| Notes: | 2. díj |
| Uncontrolled Keywords: | scarlet fever, scarlatina, count data, time-series, machine learning, poisson regression, negative binomial regression, nonlinear least squares, boosting trees, support vector machine |
| Subjects: | Mathematics. Econometrics General statistics |
| ID Code: | 10292 |
| Specialisation: | Applied Economics |
| Deposited On: | 07 Dec 2017 10:43 |
| Last Modified: | 02 Dec 2021 08:18 |
Repository Staff Only: item control page
