基于AI+EI的长期资产定价模型

Developing a Long-term Asset Pricing Model Based on AI+EI

以ChatGPT为代表的大模型激起一波新的人工智能(Artificial Intelligence,AI)热潮,资产管理行业也在寻求在各个金融场景中利用AI技术来赋能业务的可能。然而金融价格数据信噪比较低,这就对AI技术在该领域的应用提出了新的挑战。与此同时,人类专家在实际工作中可以很好地胜任多种金融分析任务,但苦于精力有限无法进一步大幅提升覆盖广度和深度。因此如何有机地整合人类专家经验(Expert Intelligence,EI)和AI技术就变成了一种有前景的工作思路,这种思路也被称为AI+EI的金融实践。本文聚焦权益资产的长期资产定价问题,提出基于AI+EI的中枢定价模型。实证表明该模型可以充分缓解资产定价当中的多样性难题,并且相比经典估值计算方法在精度上有大幅提升。

With large language models such as ChatGPT sparking a new wave of artificial intelligence(AI),the asset management industry has been actively exploring the possibility of adopting AI technologies in financial scenarios.Financial price data have a notorious low signal-to-noise ratio,which creates new challenges to utilizing AI technology in this sector.Financial experts have accumulated considerable expertise in this area and the integration of expert intelligence(EI)and AI technologies has become a promising direction for future research.This combination has become known as AI+EI financial practice.In this paper,we focus on problems with long-term asset pricing and propose a pioneering asset pricing model by adopting an AI+EI approach.The empirical results show that our model can effectively alleviate the huge diversity problem in asset pricing,and greatly improve the estimation accuracy compared with the empirical method and mainstream baselines.