How Models Work in BigQuery ML

Discover how machine learning models work within BigQuery ML through this accessible introduction to practical predictive insights. Break down complex concepts to help you understand how models fit into your data workflow and why BigQuery ML simplifies the building process directly in SQL.

The Essence of a Model

In essence, a model is a smart prediction system. It analyzes existing data, such as customer activity or sales, to learn patterns and apply them to classify new data. A model acts like a system where you feed it data, it learns, and then it predicts future outcomes.

Types of Models in BigQuery ML

Select the right model type based on your specific business questions and data structures:

• Regression models: use these when predicting a numerical outcome, such as revenue or customer lifetime value;
• Classification models: apply these to predict categories rather than numbers, such as determining if a customer will churn;
• Clustering models: utilize this unsupervised technique to identify natural groupings in data without predefined labels;
• Time series forecasting: predict future values based on past trends while considering seasonality and time-based fluctuations.

[Image comparing linear regression and logistic classification graphs]

Comparing Classification and Clustering

It is important to understand the fundamental difference between these two grouping methods:

• Classification: you work with known and predefined classes;
• Clustering: the model discovers unknown classes automatically.