It is important to document the model for transparency. To avoid unintended consequences, a second person should review the model.
- What data was used, how many columns, what is the distribution, how was data preprocessed?
- What algorithm was used, how were parameters set, what were the assumptions used?
This is important to avoid bias, irregular model generalization, evaluate model performance on unseen data. The reviewer looks at all this with a fresh pair of eyes.
How do you use this model? Let’s learn it in Model Deployment.