Professional Machine Learning Engineer

Question

You work for a large hotel chain and have been asked to help the marketing team gather predictions for a targeted marketing strategy. You must predict user lifetime value (LTV) for the next 20 days so marketing can be adjusted accordingly. The customer dataset is in BigQuery, and you are preparing the tabular data for training with AutoML Tables. The data has a time signal distributed across multiple columns. How should you ensure that AutoML fits the best model to your data?

Explanation

For time-dependent tabular data, AutoML should use a chronological split based on a designated Time column: earlier rows are used for training, later rows for validation, and the most recent rows for testing. This preserves the temporal ordering needed to evaluate predictions for future LTV and avoids leakage from future data. AutoML handles supported feature transformations, so manually combining time-related columns is unnecessary.

Learn more

Data splits for tabular data

Professional Machine Learning Engineer

AstroTutor

Notes

Readiness

QuestionQ43

Community Discussion