Методи інтелектуального аналізу даних для прийняття рішень в умовах змін на фондовому ринку

Abstract

Financial time series are characterized by non-stationarity and frequent structural changes, which reduce the stability of predictive models. The research aims to compare static and adaptive methods for forecasting price direction. The static model trained on a fixed dataset and the adaptive model using a rolling retraining mechanism. An experimental evaluation is conducted on historical stock market data and analyzed using both classification and trading performance metrics. The study evaluates the robustness of models under dynamic market conditions and provides insights into the practical implementation of adaptive learning in algorithmic trading systems.