Machine Learning-Based Financial Statement Analysis

Abstract

This paper explores the application of machine learning methods to financial statement analysis. We investigate whether a range of models in the machine learning repertoire are capable of forecasting the sign and magnitude of abnormal stock returns around earnings announcements based on financial statement data alone. We find random forests and recurrent neural networks to outperform deep neural networks and linear models such as OLS and Lasso. Using the models' predictions in an investment strategy we find that random forests dominate all other models and that non-linear methods perform relatively better for predictions of extreme market reactions, while the linear methods are relatively better in predicting moderate market reactions. Analysing the underlying economic drivers of the performance of the random forests, we find that the models select as most important predictors accounting variables commonly used to forecast free cash flows and firm characteristics that are known cross-sectional predictors of stock returns.

Keywords: financial statement analysis, fundamental value, machine learning, earnings announcement, accounting-based anomalies, prediction

JEL Classification: G12, G14, M41