What problem does it solve? Writing pandas-style data analysis and machine learning code against BigQuery requires knowing BigFrames-specific APIs, ordering modes, and ML packages, and mistakes like materializing data locally or using Scikit-learn cause out-of-memory errors and broken pipelines. ## Core Features & Use Cases - BigFrames DataFrame Best Practices: Enforces partial ordering mode, peek() previews, accessor-based transformations, and avoidance of to_pandas() and raw SQL via read_gbq(). - Cloud-Native Machine Learning: Guides model training with bigframes.bigquery.ml (linear and logistic regression) so training runs inside BigQuery instead of local memory, plus legacy bigframes.ml guidance for ARIMA Plus forecasting, PCA, and model persistence with to_gbq(). - Use Case: In a notebook, load the public penguins dataset with bpd.read_gbq(), clean it with DataFrame methods, train a logistic regression model via bbq.ml.create_model, and evaluate it without ever downloading data to the client. ## Quick Start Ask the assistant to write BigFrames code that loads a BigQuery table, cleans it with DataFrame methods, and trains a linear regression model using bigframes.bigquery.ml.
