Hands-On Machine Learning for Algorithmic Trading

Define targeted model objectives

One of the first steps in the machine learning process is the definition of an objective for the algorithm to optimize. Sometimes, the choice is simple, for example, in a regression problem. A classification task can be more difficult, for example, when we care about precision and recall. Consolidating conflicting objectives into a single metric like the F1 score helps to focus optimization efforts. We can also include conditions that need to be met rather than optimized for. We also saw that reinforcement learning is all about defining the right reward function to guide the agent's learning process.