Analýza dopravní kriminality na úrovni detailních dat PČR

This bachelor’s thesis analyses daily traffic accident counts in the Czech Republic over the period 2016–2025. The aim is to determine whether and how daily accident counts can be statistically explained by meteorological, temporal and spatial variables. The analysis draws on open data from the Police of the Czech republic (over 962,000 records), meteorological data from Open-Meteo, and spatial layers of administrative regions and the road network. A traffic crime dataset is used for contextual comparison. The methodological framework is based on generalized additive models for count data. A Poisson model is estimated first; due to detected overdispersion, a negative binomial specifi- cation is then adopted. Models include meteorological variables, calendar factors, seasonality and a COVID-19 state-of-emergency indicator. The spatial component extends the model with regional fixed effects and road class categories, and tests residual autocorrelation using Moran’s I and local LISA indicators. Results show that daily accident counts exhibit a clear systematic structure. Precipitation, snowfall, temperature, wind speed and cloud cover are all statistically significant predictors, with mostly non-linear relationships. Substantial differences are found across days of the week, public holidays and pandemic restriction periods. Spatial models confirm the dominant role of the regional factor, while residuals show no statistically significant spatial autocorrelation.