Intermediate

Snowflake ML Functions

Use Snowflake's built-in ML functions for time-series forecasting, anomaly detection, classification, and contribution analysis — all from SQL with no ML expertise required.

Overview

Snowflake ML Functions are SQL-based machine learning capabilities that let analysts and data engineers build ML models without writing Python code or managing infrastructure. They handle data preprocessing, model training, and prediction automatically.

💡
No ML expertise needed: ML Functions are designed for SQL users. You create a model with a single SQL statement, and Snowflake automatically selects algorithms, tunes hyperparameters, and validates results.

Available ML Functions

📈

Forecasting

Predict future values for time-series data with automatic seasonality detection and multi-series support.

⚠

Anomaly Detection

Identify unusual data points in time-series data using unsupervised or supervised approaches.

🛠

Classification

Build binary and multi-class classification models for categorical predictions from tabular data.

🔍

Top Insights

Automatically find the dimensions that most contribute to changes in a metric over time.

Forecasting Example

SQL Forecasting
-- Create a forecasting model
CREATE OR REPLACE SNOWFLAKE.ML.FORECAST sales_forecast(
  INPUT_DATA => SYSTEM$REFERENCE('TABLE', 'daily_sales'),
  TIMESTAMP_COLNAME => 'sale_date',
  TARGET_COLNAME => 'revenue',
  SERIES_COLNAME => 'region'  -- multi-series forecasting
);

-- Generate predictions for the next 30 days
CALL sales_forecast!FORECAST(
  FORECASTING_PERIODS => 30,
  CONFIG_OBJECT => {'prediction_interval': 0.95}
);

Anomaly Detection Example

SQL Anomaly Detection
-- Create an anomaly detection model
CREATE OR REPLACE SNOWFLAKE.ML.ANOMALY_DETECTION anomaly_model(
  INPUT_DATA => SYSTEM$REFERENCE('TABLE', 'server_metrics'),
  TIMESTAMP_COLNAME => 'metric_time',
  TARGET_COLNAME => 'cpu_usage',
  LABEL_COLNAME => ''  -- unsupervised (no labels)
);

-- Detect anomalies in new data
CALL anomaly_model!DETECT_ANOMALIES(
  INPUT_DATA => SYSTEM$REFERENCE('TABLE', 'latest_metrics'),
  TIMESTAMP_COLNAME => 'metric_time',
  TARGET_COLNAME => 'cpu_usage'
);

Classification

Build classification models for categorical outcomes:

  • Binary classification: Predict yes/no outcomes like churn prediction or fraud detection
  • Multi-class: Predict categories such as customer segments or product types
  • Automatic feature handling: Snowflake handles missing values, encoding, and feature selection
  • Evaluation metrics: Built-in accuracy, precision, recall, and AUC reporting
✅
Key takeaway: Snowflake ML Functions democratize machine learning by making it accessible through SQL. They are ideal for common ML use cases where you need quick results without the overhead of building custom ML pipelines.