| Abstract [eng] |
The work aims to construct machine learning models that recognize different regimes in stock markets and compare their results with the autoregressive Markov regime-switching model used by the Bank of Lithuania. The set of macroeconomic indicators used for training machine learning models is available on the FRED page. Metrics used to assess the accuracy of the obtained forecasts are accuracy (ACC), Brier (QPS), the area under the ROC curve (AUC) and Matthew's correlation coefficient. The regime-switching model uses macroeconomic and market indicators obtained from Bloomberg. Market regime forecasts are employed for creating investment strategies in the S\&P 500 stock index, the performance of those strategies is evaluated by calculating return and risk indicators. The findings of the study show that machine learning models trained using a set of macroeconomic indicators do not produce significantly better investment results than a regime-switching model. |
