Application of tree ensemble methods to the two-asset portfolio selection problem – a case study
Application of tree ensemble methods to the two-asset portfolio selection problem – a case study

Summary/Abstract: The aim of the study was to construct a two-asset optimal investment portfolio using machine learning and macroeconomic data at monthly and quarterly intervals. The auxiliary objective was to identify which macroeconomic variables significantly impact the estimation of the S&P 500 stock index and the USD/GBP currency pair. The framework included two steps: firstly, time series forecasts were conducted using tree ensemble methods, namely the random forest and XGBoost, and secondly, the forecasts were used as expected values to construct the portfolios. We analyze the extent to which the structure of a portfolio based on the estimated data differs from the one built using historical data. The results of the research showed that it was possible to use the macroeconomic data to efficiently forecast the considered time series and construct an optimal investment portfolio.

• Details
• Contents