A News-based Machine Learning Model for Adaptive Asset Pricing

The paper proposes a new asset pricing model - the News Embedding UMAP Selection (NEUS) model, to explain and predict the stock returns based on the financial news. The proposed model is built on top of the recent achievements in asset pricing and natural language processing. From the asset pricing perspective, the NEUS model is based on the Adaptive Multi-Factor (AMF) model proposed by Zhu et al. [2020], which provides a modern and more general framework for multi-factor models. The AMF model contains the traditional well-known Fama-French 5-factor model (FF5) Fama and French [2015] as a special case. The finance theory behind the AMF model is the Generalized Arbitrage Pricing Theory (GAPT) proposed in Jarrow and Protter [2016] and Jarrow [2016] as a modern and more general framework of the traditional Arbitrage Pricing Theory (APT) proposed by Ross [1976].