Final results with graphs are described on pages 38-41 with figures 5.1 and 5.2
Risk calculation is a fundamental part of the insurance industry. Previous research showed that the main problem of the currently used methods lies in ineffective use of available data.
The main objective of this work is to evaluate usage of artificial neural networks for estimating accident probability of motor insurance clients. Along with a grid search of the hyperparameters for the best performing neural network a comparison of different data preprocessing techniques is presented.
The results show that the use of artificial neural networks in the insurance industry is fully justified and should not be overlooked.
This thesis is written in English and is 46 pages long, including 5 chapters, 16 figures and 7 tables.