Market Strategy Choices Made by Company Using Reinforcement Learning

摘要

This work analyses the results of applying reinforcement learning tech-niques for chosing agent strategies that model the behaviour of companies within a market. Different kinds of patterns describing states of agent company and strategies describing its activities in given states were analysed. States may depend on com-pany resources (capital, stock level) and activities of customers, suppliers and com-peting companies. Strategies may be more aggressive or more conservative where the level of profit margin and the range of stock level increases are concerned.