Comparative Comparison of the Efficiency of Hybrid Model of an Agent-based & Recursive Neural Network in Automating Algorithmic Trading Strategies in Global Financial Markets

The purpose of this study is to investigate and compare the efficiency of using a hybrid model of an agent-based and recursive neural network to automate algorithmic trading strategies in global financial markets and the Tehran Stock Exchange. The model consists of two groups of agents, including traditional agents and intelligent agents. The group of traditional agents is divided into three categories: liquidity providers, liquidity consumers, and noise traders. Historical data was used to predict stock prices in the intelligent agent group, and model simulations were used to generate trading signals and update the limited order book. To extract the historical data, information from the financial markets of New York, Frankfurt, and Tokyo from 2013 to 2020 AD and the Tehran Stock Exchange from 1392 to 1399 Persian calendar was extracted from the official websites of these markets. To compare the efficiency of the model, autocorrelation and Hurst exponent tests were performed on the time series of the model price and the time series of the closing price of the historical data of the actual financial markets. The results of the autocorrelation and Hurst exponent analysis of the model and the historical financial market data were compared using the Mann-Whitney test. The results of the Mann-Whitney test show that the model can effectively predict the behavior of the actual financial markets.