BigQuery ML Tutorial: Step-by-Step Guide

  • 11/2/2024
  • One-minute read

The section below covers Quick Start Guide.

Quick Start Guide

  1. Prerequisites
    • Google Cloud account
    • BigQuery access
    • Sample dataset

Getting Started with BigQuery ML

Model Types Available

Model TypeUse CaseExample
Linear RegressionNumeric predictionHousing prices
Logistic RegressionBinary classificationCustomer churn
XGBoostComplex patternsFraud detection
Deep Neural NetworksImage/text analysisSentiment analysis

Basic Model Creation

CREATE OR REPLACE MODEL `project.dataset.churn_model`
OPTIONS(
  model_type='logistic_reg',
  input_label_cols=['churned']
) AS
SELECT
  churned,
  tenure,
  monthly_charges,
  total_charges,
  contract_type
FROM
  `project.dataset.customer_data`
WHERE
  churned IS NOT NULL;

Model Evaluation

SELECT
  *
FROM
  ML.EVALUATE(MODEL `project.dataset.churn_model`);

Best Practices

  1. Data Preparation

    • Handle missing values
    • Normalize features
    • Split train/test data

  2. Model Optimization

    • Use early stopping
    • Implement cross-validation
    • Monitor training metrics

  3. ** governance & Cost Control**

    • Track dataset access with Cloud Logging; models inherit table-level IAM.
    • Schedule model training during off-peak hours to control slot consumption.
    • Document predictive use-cases to satisfy internal model-risk-management policies.

Resources

Note: Features and syntax are current as of February 2024. Check documentation for updates.

cloudaigcpdatatutorial