Development of Multi-Agent Investment System
Anglų kursinis darbas.
Development and. Investigation of Multi-Agent Expert Investment Decision Support System. Introduction. Topicality of the research. The object of the scientific research. The objective of the work. The tasks of the work. Methodology of the research. Scientific and engineering novelty of the work. Practical significance of work results. The scope of the scientific work. Theoretical and practical aspects of securities valuation, investment decision making and portfolio formation. Theoretical and practical aspects of software agents. Formation of multi-agents expert investment decision support system. Application of multi-agents expert investment decision support system at financial markets. General conclusion. References.
Apart from fundamental and technical analyses, used in MEIDSS as the aspects of securities valuation and investment decision making, the work offered and used other forms of analysis called expert and risk analyses. In the course of investment decision making investors use advice of experts what influences their decisions. Risk analysis is intended for securities valuation taking into account the evaluations of ratings companies. The indicators included into these analyses are given in Table 1.
and possibilities to research and compare MEIDSS using different migration, communication models and scenarios. The analysis of software agents platforms helped to find a platform called JADE in which the most agent creation standards (becoming global) are implemented according to FIPA organization, and to implement both communicating and mobile multi-agent systems with a help of one tool.
Fig. 1 MEIDSS horizontal structure Fig. 2 MEIDSS vertical structure
Horizontal MEIDSS structure in intended to satisfy the needs of conservative investors and corresponds to their decision making principles based on the widest possible securities aspects evaluation round. Vertical MEIDS structure corresponds to speculative investor philosophy when decision is made taking into consideration one securities aspect, and if it generates doubts a decision is grounded taking into account other aspects chosen in successive hierarchical order. MEIDSS implementation by multi-agent system allows proving recommendations with a help of one or the other structure in comparison to other systems which usually use one structure.
MEIDSS creation (design) phase resulted in creation and implementation of the methods which allow providing recommendation for securities valuating them with a help of several aspects simultaneously. It is not a frequent event in the course of implementation of similar systems, so it can be considered as new knowledge for science and engineering.
4. APPLICATION OF MULTI-AGENTS EXPERT INVESTMENT DECISION SUPPORT SYSTEM AT FINANCIAL MARKETS
45 portfolios) they managed to outstrip the returns of adequate comparative indexes. Comparison of pseudo investor risk parameters (standard deviations and VaR) with adequate comparative index shows that investors get higher risk, however logically it corresponds to higher return, so system recommendations are considered to be correct. MEIDSS portfolio management quality was evaluated calculating Sharpe ratios, the analysis of which showed that MEIDSS portfolio management quality is not worse that it corresponds to average investment results at the market.
their combinations used for securities valuation. This research differs from the one summarized in table by the fact that investor success factor is eliminated, i.e. an investor chooses at his own and sole discretion, but not all MEIDSS recommended positions what does not allow to evaluate objectively how good securities valuation aspects operate. The results of this research are given in Table 4.
The data of Table 4 show that if securities are evaluated in all aspects in their combinations more than half of investment decisions are profitable, i.e. the ratio of profitable transactions is higher than of unprofitable ones. When considering in more detail it can be seen that in case of most aspects this ratio ranges from 50 to 75%, i.e. two of three recommended transactions will be profitable. Obtained results show that expert system implemented for securities aspects evaluation answered the purpose. Data of Table 4 allow making one more important generalization, i.e. the use of several matched aspects for the grounding of investment decision allows achieving better investment results in comparison with the use of only one securities valuation aspect.