智能差異對市場活動之影響：以遺傳規劃為基礎的代理人基雙方喊價市場模擬研究
戴中擎
摘要
本文研究目的在於探討個體認知能力對交易者在雙方喊價市場交易表現的長期影響，以及交易者的認知能力水準對市場價格造成的廣泛效果。為達此目的，本文建立一個代理人基模擬模型，並利用遺傳規劃演算法作為刻劃個別交易者認知能力的工具，以遺傳規劃中策略群體大小作為認知能力的替代參數，讓模型中的決策個體在擁有學習能力的情況下同時具有不同的認知能力。本文首先透過與真人實驗數據的比較來確認代理人的行為可以重現真人實驗的主要結果。其次，透過大規模模擬發現市場交易者認知能力對價格的影響並非像真人實驗所呈現的單一關聯性。本文發現真人實驗研究所得到結論僅在供需對稱的市場中成立，反之在供需不對稱的市場中交易者認知能力高低對價格的影響有可能呈現出非線性的關聯性。
關鍵詞：雙方喊價、代理人基模型、遺傳規劃、認知能力、學習效果
The Effects of Cognitive Ability on Market Activities: An Agent-Based Simulation Study on Double Auction Markets Based on Genetic Programming
Chung-Ching Tai
Abstract
The aim of this paper is to investigate the potential influence of traders' cognitive ability on their performance in double auction markets, and how levels of cognitive ability will impact market price in general. To do this, we use Genetic Programming to characterize the level of cognitive ability as well as learning capability of our agents in an agent-based model. The parameter of population size in Genetic Programming is regarded as a proxy parameter of cognitive ability in this study so that agents could be assigned with different levels of cognitive ability. We first compare our simulation results with human experimental outcomes to see if our model can capture the main patterns observed in human experiments. Next, we run our artificial markets with multiple simulations and find that the impact of cognitive ability on price behavior is inconsistent with previous studies. To be specific, we find that the results found in previous studies only hold when market is symmetric. The levels of cognitive ability could have a nonlinear impact on price stability.
Keywords: Double auction, Agent-based models, Genetic programming, Cognitive ability, Learning effect
Received: 2017-08-31
Accepted: 2017-09-30