||This thesis deals with applications of a new methodology of agent-based modelling on research in finance. In the first part, the methodology of agent-based computational finance is explained and some models, which use this methodology, are presented. In the next part, a model of an artificial market, the Genoa Artificial Stock Market, is introduced. We implement it in Matlab software, which enables us to assess properties of the market with emphasis on stylized facts of financial markets. Furthermore, two modifications of the model are proposed and implemented, which employ some knowledge from behavioural finance. These modifications create interdependence between returns of shares and can be one of possible explanations of financial crises. We conclude by summarizing the results and suggesting further modifications that could enhance the model and make it more realistic.