This paper examines the possibility of using evolutionary learning methods for classification. Great attention is paid to studying special features of credit assignment methods and genetic algorithm application to classification tasks. All the statements are based on the data obtained as a result of specific task solving: lepiota and agaricus family mushroom division into edible and poisonous. To implement the system, special software was developed in C++. As a result, a system was constructed able to classify approximately 90% of the mushroom varieties suggested. After a slight modification of the algorithm and optimisation of the parameters, the system was able to produce about 95% of correct answers.