05. Model Selection: Choosing the Right Algorithm
PRTDM2-785 AI Trading C2 L1 Vid6 Model Selection
Choosing the Right Algorithm
Selecting an algorithm involves a blend of science and art, aiming to find the best-suited model from several viable options based on different criteria:
Output Type: Ensure the algorithm aligns with your desired outcome.
- Supervised Learning: Algorithms learn from labeled data to predict outputs for new data, like predicting house prices based on historical sales.
- Unsupervised Learning: Deals with unlabeled data, utilizing clustering to categorize information like customer purchase habits.
Algorithm Assumptions: Respect the model's assumptions to prevent sub-optimal results. For instance, violating a linear regression's linearity assumption can lead to inaccurate predictions.
Package Availability: Algorithms must run on the system's available packages to be useful.
Data Volume: Match algorithm requirements with data scale; some need vast data, particularly deep learning models not covered here.
Start with simple models and expand complexity as skills improve.