Skutar I.D., Yurchenko I.V., Yurchenko O.I. Hybrid System For Analysis And Forecasting Of Financial Asset Dynamics Based On Stochastic Differential Equations And Recurrent Neural Networks // Scientific World Journal.– 2026.– Issue №36, Part 2.– P.152–164.– Published by Academy of Economics named after D.A. Tsenov, Bulgaria (jointly with SWorld).

Abstract

This article addresses the problem of developing hybrid information-analytical systems for financial markets capable of operating in real-time. It proposes the architecture of a web-based application that integrates classical stochastic calculus models (Geometric Brownian Motion and the Vasicek model) with deep learning algorithms (LSTM recurrent neural networks). The study describes the mathematical framework, numerical discretization methods (Euler-Maruyama scheme), and market noise generation techniques (Box-Muller transform). The choice of the technology stack (React.js, WebSocket) and streaming data management algorithms (sliding window) is justified. Finally, an analysis of the obtained simulation results and the accuracy of the neural network forecasting is conducted.