Towards a multi-agent based architecture to simulate the reality of a stock exchange market

Abstract

In this paper, the initial steps of the development process of an agent-based architecture used to simulate the reality of a stock exchange market are provided. This architecture includes a representation of all market participants from investors to floor traders. There are five main modules in the proposed architecture: an information retrieval and ontology module which extracts online market news and data, a knowledge base which includes portfolios compositions, a strategy and planning module used as an analysis tool, a learning module based on market historical data where agents learn market behavior and tendencies, and finally a decision making module with which agents make decisions on buying and selling of stocks. The information retrieval and ontology module is then detailed to show the proposed ontology representing the financial news. This architecture can be used to develop simulation platforms for stock markets.