Predicting the ranking of the sales surprises in the context of active investing

Details

Supervisors

Vande Kerckhove, Corentin

Faculty

Louvain School of Management

Degree label

Master [60] en sciences de gestion (horaire décalé)

Abstract

Quantitative investing refers to the action of trying to optimize the return of an actively managed portfolio on the stock market. It has been shown that using pure machine learning models to forecast the return of stock prices directly is a very hard task. Different mechanism have been tried to circumvent that issue. This thesis focuses on the forecasting of fundamental values. More specifically, we are forecasting the surprise on the consensus of a given fundamental value. Indeed, it has been shown that this surprise can be used to create high returns investment strategies. A few papers look at forecasting the surprise directly, even if what matters in the end in the investing heuristic that we use is the ranking of several surprises on a set of companies. Therefore, in this thesis, we try to use ranking-specific objective functions for machine learning models that try to rank the surprises on a set of given companies. We show that some improvement on the metric can be obtained using this strategy, leading to potential higher returns on the portfolio.