Multi–agent Environment for Modelling and Analysing Market Strategies
Paweł Kolarz, Jan Marszałek, Jarosław Koźlak, Małgorzata Żabińska · Procedia Computer Science · 2012
Our goal was to develop an environment which will make it possible to model and optimise supply chains and company behaviours, as well as to test the influence of applying different decision algorithms. The system consists of a set of companies which cooperate and compete with one another by offering and buying products and negotiating their conditions of transactions. Throughout the presented work we focused on the realisation of the parts of the strategic planner module. Producers tend to maximise profits, among other ways, by modifying the margin of offered products.The main aim of the applied decision algorithm is to have a feature of adapting to a given situation: to select the best model of the simulation world and to choose the most adequate actions in the given situations. The first aspect is addressed by the use of an adaptive algorithm which chooses the best demand prediction algorithm in a given situation. The second aspect is addressed by the clustering technique which identifies the similar situations on the market as one possible state and assigns with it the most profitable action.