Authorship identification in the context of this thesis is a classification task that assigns a human-written digital text to an author from a known set of authors. There are many different state-of-the-art algorithms for classifying authors of text based on different classification algorithms and text processing techniques. However, papers proposing these solutions often provide their evaluation only on one selected dataset. The research question is how robust the different algorithms on datasets of different types of text (e.g., emails, articles, or instant messages) are. The aim of the thesis is to implement three selected algorithms and compare their performance on two or three selected datasets. The resulting evaluation will indicate which of these algorithms has the best performance regardless of the nature of the dataset. The thesis is done within the Natural Language Processing Centre (NLPC) in cooperation with Konica Minolta.