Question

A subscription media company wants to predict customer churn using two years of historical account data stored in BigQuery. The data science team is small and lacks deep expertise in building and tuning custom neural network architectures, but they need an accurate classification model deployed as an online prediction endpoint within a few weeks. They want to minimize the amount of custom modeling code they write while still leveraging their structured tabular data. Which approach on the Gemini Enterprise Agent / Vertex AI platform best meets these requirements?

Accepted Answer

Use Vertex AI AutoML for tabular data to train a classification model directly from the BigQuery dataset, then deploy it to a Vertex AI online prediction endpoint.

Answer

Write a custom TensorFlow model, containerize it, and run distributed training on Vertex AI custom training jobs before deploying to an endpoint.

Answer

Call the prebuilt Natural Language API to score each customer record and aggregate the sentiment results to infer churn likelihood.

Answer

Export the BigQuery data to Cloud Storage and use the Vision API to classify customer behavior patterns into churn categories.