A REVIEW OF ADVANCEMENTS AND APPLICATION OF THE STOCK MARKET IN Q-LEARNING REWARD FUNCTION ANALYSIS

Abstract

Deep Q-learning is a common reinforcement learning approach in this technique including training of this method using deep neural networks (DNN), in this technique also using Deep Q-networks (DQN) network approximates widely recognized Q-functions. The popular version of deep Q-learning is developed under practical and verifiable principles by a dynamical viewpoint, which also offers a theoretical analysis, despite the lack of formal guarantees that would limit its usage in reality. Algorithms are analyses of convergence of the asymmetric behavior of the educational process. Stock exchange forecasting is now in high demand, and most investors predict challenging tasks, researcher’s analysts of the financial market are noisy, volatile, nonparametric, complex, nonlinear. Technology used Deep reinforcement learning (DRL) algorithms previously to find intractable problems. DRL makes it possible to generate automation profit in the stock market financial combining price assets prediction steps and allocation steps used in this portfolio method is a unified process is a fully autonomous system capable of environment interaction to make the decisions optimal through fault and trial. The paper focuses on showing various reward functions used for learning patterns and making optimum decisions.