Удаление противоречивых наблюдений как процедура предрегуляризации нейросетевой модели налогового контроля

Stanislav Anatolevich Gorbatkov; Ilya Igorevich Belolipcev; Elena Anatolevna Murzina

Removal of conflicting data as a procedure of regularization of neural network model of tax control

Authors

Stanislav Anatolevich Gorbatkov
Ilya Igorevich Belolipcev
Elena Anatolevna Murzina

Keywords:

Conflicting data; regularization of training; sustainability neural network model.

Abstract

An algorithm for finding and removing conflicting observations from the data is proposed. The proposed procedure is a procedure of regularization neural network model of tax control and is aimed at improving stability of the model to a perturbation of the input data.