| Summary: | 個人投資理財是近年來熱門的議題，而國內與此相關的研究大都集中在投資績效的提升、投資標的之選擇、資產配置比例與影響投資績效的變數等，較少在探討個人的投資準則與社會關係學習所造成的投資績效差異。本論文利用代理人為基礎的模擬方式（Agent-Based Simulation）與動態虛擬社會關係，模擬共同基金投資市場的交易行為，讓模型中的一個或數個群體裡的所有投資人除了可以依循著自身的投資準則進行投資外，亦有機會藉由虛擬社會關係學習到其他投資人的投資準則，進而提升投資績效。在實驗中，我們針對不同的學習頻率及學習參數觀察學習的效果。我們發現，當有虛擬社會關係學習模式且學習評估頻率為每月一次時，有助於整體投資績效的提升。 === Personal investment is a topic that has attracted much attention in recent years. However, the researches and applications related to this topic are usually concentrated in the area of increase investment performance, portfolio, investment selection, and critical investment performance factors. Less are about investment criteria and social learning that affect investment performance.
In this thesis, we use agent-based simulation with dynamic virtual social relationship to simulate artificial mutual fund market. The investors in the model can invest by their own criteria, and learn other agent’s criteria via virtual social relationship to increase investment performance. We use different sets of parameters in the experiments to observe how these parameters affect the result. Our experiments revealed that our new model with social learning mechanism and a learning evaluation frequency of a month, the overall investment performance can be significantly improved.
|
|---|
