INFORMATION SUPPORT FOR DECISION-MAKING USING DISTRIBUTED ARTIFICIAL INTELLIGENCE TECHNOLOGIES

Abstract

The methodology of this study is based on the assessment of theoretical developments and practical use of distributed artificial intelligence (AI) technologies in decision-making activities. The emphasis is on the implementation of machine learning methods for analysing large amounts of data in distributed systems, which allows for efficient and prompt decision-making support. Tools and approaches used Python and multiprocessing to build parallel data processing. Logistic regression is used to describe decision-making processes and to assess the classification and accuracy of the resulting models. PCA – principal component analysis to reduce the dimensionality of data for clustering and classification purposes. Privacy control methods to ensure the security of data transmission between nodes and restrict access to data. Methodology. The methodology of this study is based on the assessment of theoretical developments and practical use of distributed artificial intelligence (AI) technologies in decision-making activities. The emphasis is on the implementation of machine learning methods for analysing large amounts of data in distributed systems, which allows for efficient and effective decision-making support. Tools and approaches used: Python and multiprocessing to build parallel data processing, logistic regression to describe decision-making processes and evaluate the classification and accuracy of the resulting models, PCA – principal component analysis to reduce the dimensionality of data for clustering and classification purposes, privacy control methods to ensure the security of data transmission between nodes and restrict access to data. Scientific novelty. The main material emphasises the advantages of distributed AI, both functional (data processing speed, parallelism, and high scalability, technical feasibility) and development (lower costs, reusability, and implementation flexibility). It also takes on several challenges, such as AI liability, data privacy and security issues, and implementation issues. It has been found that distributed AI is an immensely powerful technology and care should be taken when applying it to decision-making, as while it has advantages, it also has some disadvantages. In this paper, we demonstrate how AI systems running in distributed architectures can be used for decision making, using our own Python development as an example. We describe such a system that uses logical regression, principal component analysis, and visualisation of results. The article focuses on finding a solution by executing Python code developed in the PyCharm 2024.2 integrated development environment, which demonstrates a distributed AI-based decision-making system using multiprocessing and machine learning methods. The code includes functions for training models on individual nodes, managing the distributed learning process, and visualising the decision boundary of the best model. It uses logistic regression for binary classification and PCA for dimensionality reduction to facilitate visualisation. The code creates a pool of processes that correspond to a given number of nodes. These processes then train the models asynchronously on subsets of the data. After the training process is complete, the best model is selected, and the decision boundary is represented as a two-dimensional graph in the simplest case. The distributed AI described in this example embodies a potential application of the technology in decision-making: the ability to make decisions over distributed data and distributed actors by processing data on multiple nodes. Conclusions. It can be argued that distributed AI technologies can be integrated into decision-making processes in any case. It is possible that organisations would like to rely on a culture of critical evaluation and continuous learning to make more informed, fair, and effective decisions. The code developed here is an example of how such a system could be implemented in practice.