採用由下而上的模擬方法, 與大量交涉溝通的代理人來建構財務市場的模型，已開始展露頭角並成為另一研究的方法。財務市場是以代理人為基礎來做模擬的重要應用，聖塔菲研究院 (Santa Fe Institute)所提出的虛擬股票市場(Artificial Stock Market)正是此類重要的模型應用，並為科學家研究的重要標竿之一。然而其一設計者Blake LeBaron指出，此虛擬市場內的交易者無法達到規則的溝通交換，需要整合社會學習的機制。基於此原因，我們利用Java-Swarm ASM 2.2為虛擬股市平台，在其上面建立預測規則交換的機制來解決此問題。我們採用的學習平台能使交易者發佈其表現良好的規則給其他人使用，或是從平台上取得其他績效不錯的規則來使用。在這種方式下，社會學習行為例如投資者的群聚、消息的散播等，都能在此市場上模擬出來。然後我們設計多個實驗來模擬真實股票市場的運作，我們收集時間序列資料上的巨觀性質，並測試此加上社會學習機制的市場是否存有唯一對稱的Nash 平衡 (symmetric Nash equilibrium). 我們藉由這些資訊觀察出僅有個人學習行為與擁有社會學習行為市場的不同。我們証明加入社會學習行為機制後的市場是合理的，且與真實市場行為更加近似。 Modeling financial markets from the bottom up with large numbers of interacting agents is beginning to show promise as a research methodology. Financial markets are an important application for agent based modeling styles. Santa Fe Institute artificial stock market is a well known agent-based model and one of the benchmarks for researchers to study. Nevertheless one of its designers, Blake LeBaron, brought up a drawback that this market lacked of rule sharing and needed to coordinate social learning between agents. On the basis of this reason, we built a rule sharing platform on Java-Swarm ASM 2.2 to solve this problem. We utilized a leaning pool to make agents having the capability of social leaning. Agents who are publishers may publish rules into learning pool to share with others; receivers may retrieve better rules from learning pool. In this way, social learning behavior such as herd behavior or rumors dissemination is simulated in this market. Then we made several experiments to simulate how real markets operate. We gathered macro time series data, and tested whether there is a unique symmetric Nash equilibrium in our modified market. We observed the difference between markets with social learning and individual learning mechanism, and we proved that our modified SFI-ASM is a rational one compared with realistic market. We can observe more features common to real markets in this modified market.
