UDC 519.7
DOI: 10.36871/2618-9976.2023.12.007
Authors
Nikita A. Andriyanov,
Candidate of Technical Sciences, Associate Professor of the Department of Data Analysis and Machine
Learning, Financial University under the Government of the Russian Federation, Moscow, Russia
David A. Petrosov,
Candidate of Technical Sciences, Associate Professor, Associate Professor of the Department of Data
Analysis and Machine Learning, Financial University under the Government of the Russian Federation,
Moscow, Russia
Andrey V. Polyakov,
Graduate Student of the Department of Data Analysis and Machine Learning, Federal State Budgetary
Educational Institution of Higher Education Financial University under the Government of the Russian
Federation, Moscow, Russia
Abstract
This article proposes a study aimed at determining the architecture of artificial neural networks to solve the problem of determining the population state of a genetic algorithm adapted to solve the problem of structuralparametric synthesis of simulation models of business processes. As the initial data for training the artificial neural network, we used the results of computational experiments obtained when operating a genetic algorithm model based on mathematical nested Petri nets, which solves the problem of synthesizing business process models (Petri net models) based on a given behavior. As examples of artificial neural network architectures for managing the process of finding solutions based on an evolutionary procedure, the following are considered: fully connected artificial neural network (FCNN), simple recurrent artificial neural network (Simple RNN), long shortterm memory recurrent network (LSTN), closed recurrent recurrent network block (GRU) and bidirectional LSTM (Bidirectional LSTM). The deep learning algorithms used were: Support Vector Classifier, Decision Tree Classifier and Random Forest Classifier. The article discusses the presented architectures of artificial neural networks and various training methods. Based on the computational experiments carried out and the analysis of the results obtained, conclusions were drawn about the feasibility of using artificial neural networks with RNN architecture to solve the problem of recognizing the state of the population and controlling the process of synthesis of solutions.
Keywords
Artificial neural networks, Structuralparametric synthesis, Business processes, Intelligent systems, Genetic algorithm, Machine learning, Deep learning, Artificial neural network
