Internet and other new media make it easier for private investors to trade on the stock markets. Unfortunately, a lot of them end up with losses and bad experiences. The major causes of this problem are the lack of technical know-how about stock market mechanisms, the fact that people have little time and discipline to follow up their investments and finally because the amount of data at the disposal of the private investors is overwhelming.
The major goal of this masters thesis is to develop an integrated software system to support the private investor in making the right decisions at the right time. Existing commercial software for this purpose is restricted to technical analysis of stock data and as a consequence the user should have a lot of knowledge about these techniques. Another disadvantage of existing software is its passive nature: the user is the only one to take initiative and there is no autonomous decision making to propose particular investments.
To overcome these disadvantages, we have developed a system using an agent-based software approach. This implies a software system with a high level of autonomy. We propose an agent-architecture composed of several subsystems. A first subsystem is responsible for gathering and managing financial data of various kinds: stock quotes, financial reports of companies quoted on the stock exchange, macro-economic trend indicators, etc. A second subsystem is an extensible library of building blocks to compose various algorithms for information processing and interpretation, in order to extract relevant financial signals. The third subsystem, the software agent itself, is responsible for the autonomous behavior of the entire system. It makes use of the financial data and various algorithms for data analysis to come up with specific proposals for financial transactions. These results are used to advise the investor, or one could allow the system to place the proposed orders automatically.
One of the most important aims during the development of the software system was the aspect of flexibility. It is absolutely necessary that every part of the system can be configured according to the specific needs of a particular user: algorithms for the analysis and interpretation of financial data can be composed and parameterized at will, the strategy to be used by the agent can be adapted to the wishes of each investor (e.g. defensive or aggressive strategy according to the risk the user is willing to take). In order to control all this flexibility a configuration mechanism based on a meta-language, defined in XML, was developed. It is possible to configure each important aspect of the system using a single XML file.
As an evaluation, we did some simulations with the electronic stock market agent using various configurations and a completely autonomous execution without human interaction. Although the composition of strategies with high return rates was not a goal of our work, the extra financial benefits achieved by the iron discipline and the accurate task execution of the electronic stock market agent are certainly worth mentioning.