The purpose of this thesis is to discover typical, frequent episodes using the Minimum Description Length principle.
In the first part of this paper, there is a short description of the actuality and importance of the topic. After that the MDL and other methods are presented. The evolution and the variants of the minimum description length princple are explained in details. The MDL method is compared to the Minimum Message Length method too.
The second part of the paper deals with implemetation. The created method follows the minimum description length principle. With it, the typical episodes are discovered in an event sequence. The importance of repeating episodes often can be characterized by how much they can reduce the description of the whole data. Beside the detailed description of the implemented method, other algorithms used for compression are described. These are used for comparison.
In the last part of the paper the achieved results and ways for further improvements are stated.