Enabling model retraining

So far, we've talked about what model drift is and how to recognize it. So, the question is, what should we do about it? If a model's predictive performance has deteriorated due to changes in the environment, the solution is to retrain the model using a new training set that represents the current situation. How much should your model be retrained by? And how can you choose your new workout routine? The following diagram shows the Model retrain function triggering the Build module based on the results of the Monitor module. There are two ways to trigger the model retrain function. One is manually and the other is by automating the model retraining function. Let's see how we can enable both:

Figure 13.17 – Model retraining enabled in an MLOps workflow

Manual model retraining

The product owner or quality assurance manager has the responsibility of ensuring manual model retraining is successful. The manual model...