Stock Prediction with ML: Ensemble Modeling -- The Alpha Scientist

Markets are, in my view, mostly random. Many small inefficiencies and patterns exist in markets which can be identified and used to gain slight edge on the market. These edges are rarely large enough to trade in isolation - transaction costs and overhead can easily exceed the expected profits offered. But when we are able to combine many such small edges together, the rewards can be great. In this article, I'll present a framework for blending together outputs from multiple models using a type of ensemble modeling known as stacked generalization. This approach excels at creating models which "generalize" well to unknown future data, making them an excellent choice for the financial domain, where overfitting to past data is a major challenge.