.md .pdf

Black Swan

work in progress

• Potentially, complete rebaselining

• Deliberately degrade the model’s accuracy over time. Why? Shift in attributes, and predictions.

• Introduce data drift, and concept drift as concepts.

• This is where the proxy to COVID may be relevant.

• https://www.bankofengland.co.uk/bank-overground/2021/how-has-covid-affected-the-performance-of-machine-learning-models-used-by-uk-banks

• https://medium.com/eliiza-ai/why-your-models-might-not-work-after-covid-19-a00509e4920b

• https://towardsdatascience.com/the-covid-19-concept-drift-using-sydney-ferry-activity-data-32bbff63cb9f

• How do we handle these situations?

  • Establish a new pipeline, with data after xyz date and train a new model

  • In practice, are the changes this dramatic?

previous

Concept Drift

next

Solution Costs