Vertex AI – model monitoring
After model deployment, we need to monitor it since the data and environment may change and cause the model to deteriorate over time. Two concepts of monitoring should be considered: feature skew and drift detection.
In our demo documentation, we are going to build a brand-new tabular dataset and train the model. In this example, we will be using the Women’s International Football Results (https://www.kaggle.com/datasets/martj42/womens-international-football-results) dataset.
We have created a tabular dataset where we have uploaded a CSV file that’s been downloaded from Kaggle. The following screenshot shows a summary of the dataset:
We have also trained a model using the AutoML method, and as the target, we have used the neutral column, which has two values (either False or True). The following screenshot shows the summary of the trained model:
With Explainable AI, we...