Machine learning for financial forecasting, planning and analysis: recent developments and pitfalls

This article is an introduction to machine learning for financial forecasting, planning and analysis (FP&A). Machine learning appears well suited to support FP&A with the highly automated extraction of information from large amounts of data. However, because most traditional machine learning techniques focus on forecasting (prediction), we discuss the particular care that must be taken to avoid the pitfalls of using them for planning and resource allocation (causal inference). While the naive application of machine learning usually fails in this context, the recently developed double machine learning framework can address causal questions of interest. We review the current literature on machine learning in FP&A and illustrate in a simulation study how machine learning can be used for both forecasting and planning. We also investigate how forecasting and planning improve as the number of data points increases.

How to gauge investor behavior? A comparison of online investor sentiment measures

Given the increasing interest in and the growing number of publicly available methods to estimate investor sentiment from social media platforms, researchers and practitioners alike are facing one crucial question – which is best to gauge investor sentiment? We compare the performance of daily investor sentiment measures estimated from Twitter and StockTwits short messages by publicly available dictionary and machine learning based methods for a large sample of stocks. To determine their relevance for financial applications, these investor sentiment measures are compared by their effects on the cross-section of stocks (i) within a Fama and MacBeth (J Polit Econ 81:607–636, 1973) regression framework applied to a measure of retail investors’ order imbalances and (ii) by their ability to forecast abnormal returns in a model-free portfolio sorting exercise. Interestingly, we find that investor sentiment measures based on finance-specific dictionaries do not only have a greater impact on retail investors’ order imbalances than measures based on machine learning approaches, but also perform very well compared to the latter in our asset pricing application.