Application of neural networks to forecast the number of road accidents in provinces in Poland.

ABSTRACT

Many people die on the streets every year. Year after year this number is decreasing, but there are still a lot of them. Although COVID-19 has reduced the number of traffic accidents, this figure is still very high. For this reason, in order to identify the federal states with the highest number of traffic accidents and to do everything possible to minimize the analytical value and improve road safety, we will develop accident forecasts for the next few years. Need to know. The author's aim is to predict the number of road accidents by state in Poland, but this has not been done for many years. For this purpose, monthly data from police statistics on the number of traffic accidents by state in Poland were analyzed. Based on this data, a forecast of the number of traffic accidents in the next years from 2022 to 2024 was created in Statistica. A selected neural network model was used to predict the number of traffic accidents. The results show that a reduction in the number of traffic accidents on Polish roads can still be expected, but the prevalent COVID-19 confounds the results obtained. The choice of number of samples (training, testing, and validation) affects the results obtained.