Results of research in the field of methods of classification of badly structured collections of heterogeneous documents are provided. The method is offered and the hybrid algorithm of classification of text documents is developed. The theoretical assessment of complexity of the offered method is executed and the pilot study is made. In article it is considered: mathematical model of the offered hybrid method, its algorithm, theoretical assessment of computing complexity and the experimental comparing with other methods of classification of documents.
Keywords: fuzzy logic, bayesian classification, classification of text documents.