「資產配置」即是實踐了投資分散的原則,藉由將資金分散投資於不同的資本市場,來降低投資的風險。傳統上,以基於報酬 / 風險為基礎Markowitz E-V模式做為資產配置的方法。評價報酬 / 風險使用兩種方式:直接評價與多情境法。在使用多情境法時,必須面臨大量複雜的資料與各種不同的財務分析模式,增加了制訂投資決策的困難。智慧代理人系統藉由許多代理人合作完成任務的運作模式,產生很大的彈性,可解決使用多情境資產配置模式,所遭遇到大量財務資料與模式的困境。本研究基於智慧代理人系統架構,定義了Investor Agent、Asset Allocation Decision Support Agent、Scenario Define Agent、Scenario Probability Forecasting Agent、Asset Evaluation Agent、Financial Data Collection Agent等,模擬一個以多情境資產配置模式為基礎的資產配置決策環境。
Asset allocation is the choice of how much to invest among classes of assets to achieve the best portfolio given the investor’s objectives and investment constraints. The approach that explores the nature of risk-return trade-off and the principles of rational portfolio selection associated with it was first proposed by the Markowitz Model. In practice, the number of required data inputs for applying Markowitz model with multi-scenario analysis is extremely large. In addition, the sophisticated valuation models and computations complicate the situation. We therefore proposed an intelligent agent system for multi-scenario asset allocation decision support. By way of many intelligent agents’ co-work and coordination, the large amount of data computation and various jobs of analysis can be flexibly performed. In this thesis, we defined the roles and jobs of different agents, including: financial data collection agent, scenario-defined agent, scenario probability forecasting agent, investor agent, asset evaluation agent, and asset allocation decision support agent, to simulate multi-scenario asset allocation decisions.