MATHEMATICAL MODEL OF AN ARTIFICIAL NEURAL NETWORK FOR SOLVING DATA MINING PROBLEMS

Abstract

Background. The article discusses a neural network (artificial neural network) as a kind of mathematical model. Also, the work analyzes its software and hardware implementation. Materials and methods. The neural network method is associated with deep learning. The proposed model is built on the principle of organization and functioning of biological neural networks – networks of nerve cells of a living organism. It is a system of interconnected and interacting simple processors in the form of artificial neurons. When connected in a large network with controlled interactions, these simple processors taken separately are capable of performing quite complex tasks together. Results. As a result of the research carried out, ensemble methods can be noted, which are a method of intellectual learning, where several models are trained to solve a single question posed and are combined to obtain the best results. The main assumption of the application of the method: with the right combination of weak models, more reliable and accurate results can be achieved. Conclusions. The described ensemble machine learning methods are so-called metaalgorithms that combine several machine learning methods into one predictive model. These algorithms consist of two steps: creating a distribution of simple ML models over subsets of the original data and combining the distribution into one "aggregated" model.

Keywords

• neural network
• mathematical model
• hardware implementation
• processor
• tasks