In Search of Return Predictability: Application of Machine Learning Algorithms in Tactical Allocation

In this chapter, the authors investigate the use of machine learning algorithms in the context of market timing strategies aiming at exploiting return predictability, in order to outperform the buy-and-hold strategy. They use publicly available market data and open-source software to evaluate the usefulness of several machine learning algorithms for tactical asset allocation between bonds and equities. The authors describe the analyzed data followed by a section explaining the tactical asset allocation framework. They provide an overview of the investigated machine learning algorithms with the corresponding evaluation criteria. The authors discuss the results of the implemented strategy followed by the shortcomings of research and concluding remarks. They describes the data sources and investment problem facing the investor. The authors also provide an overview of several machine learning algorithms implemented in their empirical investigation. The tuning process implemented to estimate the hyperparameters needed for algorithms is also examined.