Question about Training Features and Fairness of Baselines in Table 1 #25
Description
Hello, I have a few questions about your work AlphaForge that I would like to ask: In the main experiment of Table1, what are the training features of the three baselines for machine learning? Are they ohlcv or some factor library? The "Machine Learning-based Alpha Factors" section of the Related Work chapter describes the Machine Learning baseline by saying: "Early approaches... merely using the historical time series data of each stock to predict stock prices, including... LightGBM... and XGBoost. Does this mean that the baselines for these three types of machine learning use raw data rather than factor libraries as training features? Is there any control that the number of factors used by different methods in table1 is the same? If there is no control, how can it be proved that alphaforge is more effective than other methods? Thank you. Looking forward to your reply!!