Third Party Machine Learning Pipeline: Leveraging the power of Scikit-Learn

---

4.5 Should feature selection be part of the Machine Learning automated pipeline?

Feature selection in CI/CD 장단점

Advantages

• Reudced overhead in the implementation of the new model
• THe new model is almost imediately available to the business systems

Disadvantages

• Lack of data versatility
• No additional data can be fed through the pipeline, as the entire processes are based on the first dataset on which it was built

Feature selection in CI/CD

Suitable:

• Model build and refresh on same data
• Model build and refresh on smaller datasets

Not Suitable:

• If model is built using datasets with a high feature space
• If model is constantly enriched with new data sources