This paper discusses statistical methods, as well as machine learning methods for choosing the optimal way to establish authorship for a passage of a work. The authors create a dataset from the passages of the corresponding authors, create a set of numerical features corresponding to each passage and apply various approaches to analyze authorship, such as correlation, similarity, t-test. An attempt is made to find the optimal method for the output layer of a graph convolutional neural network used for data preprocessing. The GCN neural network is being trained.

Keywords: t-test, cosine similarity, correlation, graph convolutional neural networks, natural language analysis