Machine Learning Methods in Asset Pricing

Abstract

This paper evaluates the traditional asset pricing models and examines the literature on the most promising machine learning techniques that can be used to price securities. Asset price forecasting is essential to efficient markets. Capital Asset Pricing Models (CAPM), Arbitrage Pricing Theory (APT) and a multitude of Factor Models are used to price securities and to establish mean variance optimal portfolios. An increasing number of scholars and financial practitioners have begun to explore the role of machine learning in asset pricing. We show how these methods have been applied in academia and discuss their results in maximizing the Sharpe Ratio. We also explore the potential use of neural networks in asset pricing. We believe that their capacity to process large amounts of data and their ability to accurately capture non-linear relationships makes them a useful estimation tool.