Analýza dopravní nehodovosti na datech Policie ČR

of knowledge discovery in databases. The work is motivated by the societal importance of road accidents and the opportunity to make use of an extensive dataset drawn from the records of the Czech Police for the years 2010–2025. The dataset captures approximately 1.5 million accidents and 1.8 million participating drivers. The theoretical part summarises the CRISP-DM methodology, association rules and the GUHA approach, and provides a more detailed treatment of the 4ft-Miner and CF-Miner procedures implemented in the CleverMiner library. The practical part covers data preparation, exploratory analysis and four modelling tasks. Two accident-level tasks search for combinations of circumstances associated with injury and with severe consequences. The driver-level task focuses on driver characteristics, and the fourth task uses the CF-Miner procedure to describe profiles of accident severity categories. The output of the thesis is a set of interpreted rules capturing risk combinations of accident circumstances and an assessment of which types of attributes carry the strongest signal with respect to the severity of consequences.