Agent based modelling of a single-stock market on the JSE

Abstract

The application of agent based modelling in nance allows market experiments to be undertaken which would normally be prohibitive due to cost, complexity and other factors. Agent based models use simple behaviour and interaction to produce complex outcomes. We introduce the requirements of an agent based market simulator based on protocol stipulated by the Johannesburg Stock Exchange. The requirements are then translated into a technical design. This design is implemented using the Microsoft .NET framework. The product of this design and creation approach is a market simulator which is then used to run three simulations where different agent behaviour is demonstrated. The approach and results of the simulations are documented to show possible use cases of the simulator.