University of Natural Resources and Life Sciences, Vienna (BOKU) - Research portal

Logo BOKU Resarch Portal

Gewählte Master / Diploma Thesis:

Christina Leopoldseder (2018): Prognose der jährlich bei Straßenverkehrsunfällen getöteten Personen anhand von vorläufigen Wochenergebnissen.
Master / Diploma Thesis - Institut für Verkehrswesen (IVe), BOKU-Universität für Bodenkultur, pp 98. UB BOKU obvsg FullText

Data Source: ZID Abstracts
In general, the statistics about traffic and road accidents is published a few months after the end of a year. At this time, the development in this sector can only be evaluated afterwards. But the Austrian Ministry of Interior already levies the number of people deceased through a car accident in the past week and the cumulative number over the year every week. With this data, an estimation over the process in this sector is already possible in the first months of a year. Though, a valid calculation model for this estimation or prediction is not known. In the following master thesis, two different models for this problem were developed. The first model depends on a Log-Linear-Poisson Regression and predicts the expectable number of road fatalities for every week of the following year. Afterwards, this number is summed up, to enable a comparison to the actual number. The second model depends on a multiple linear regression and predicts the share of the number of road fatalities in a week referred to the number at the end of the year. With this model, the prediction changes every week, according to the actual number of people deceased trough a car accident. As both models are based on a different approach, an immediate comparison of the forecast accuracy is not possible. Whereas model 1 compiles a prediction for further years, depending on the development of recent years, model 2 predicts the further process of a year, depending on the previous development. Therefore, it is recommended to work with both models in the future. Model 1 is necessary to evaluate, if the actual numbers are in accordance with the recent trend and model 2 can be used to estimate the expected number for the end of a year as early as possible.

Beurteilende(r): Berger Wolfgang Josef
1.Mitwirkender: Hössinger Reinhard

© BOKU Wien Imprint