Hybrid stock analysis model for financial market forecasting

Abstract

Various approaches are used to analyze stocks for the purpose of forecasting the financial market. Because stocks exist in a large and interconnected market, traditional methods based on time series information for a single stock do not take into account the relationships between other stocks. Taking into account the relationships between stocks can improve the effectiveness of stock price forecasting. The paper proposes a hybrid stock analysis model that uses a combination of various intelligent technologies: recurrent neural networks (RNN), artificial immune systems (AIS), and graphical neural networks (GNN). Time series in the form of daily sales volumes and stock prices are fed to the inputs of the RNN to obtain stock price characteristics. These characteristics are fed to the input of the clustering model to obtain information about the relationship between stocks in the form of a graph with selected clusters of stocks. The GNN inputs are a graph whose nodes display the characteristics of a stock exchange time series, and the arcs show the connectivity between them. The outputs of GNN are stock returns. Using this model allows you to more effectively predict the financial market and make more informed decisions in order to obtain high profits with low risks.