TACTICAL-ASSET-ALLOCATION-AND-MACHINE-LEARNING
This paper studies how a machine learning algorithm can generate tactical allocation which outperforms returns for a pre-defined benchmark. We use three distinct and diverse data sets to implement the model which tries to forecast the next monthβs a selected equity index price. The algorithm used to accomplish this task is Elastic Net. Once the predictions are generated from an out-of-sample subset, we elaborate a tactical portfolio allocation aiming to maximize the return of a different combination of classical allocation between bonds and equity, and a risk parity strategy. Finally, we evaluate those returns by comparing them to the benchmark.