Regression

NeuronDB provides linear and non-linear regression algorithms.

Linear Regression

Copy
CREATE TEMP TABLE linreg_model AS
SELECT neurondb.train(
    'default',
    'linear_regression',
    'training_table',
    'target',
    ARRAY['features'],
    '{}'::jsonb
)::integer AS model_id;

Ridge Regression

Regularized linear regression with L2 penalty:

Copy
CREATE TEMP TABLE ridge_model AS
SELECT neurondb.train(
    'default',
    'ridge',
    'training_table',
    'target',
    ARRAY['features'],
    '{"alpha": 0.1}'::jsonb
)::integer AS model_id;

Lasso Regression

Regularized linear regression with L1 penalty:

Copy
CREATE TEMP TABLE lasso_model AS
SELECT neurondb.train(
    'default',
    'lasso',
    'training_table',
    'target',
    ARRAY['features'],
    '{"alpha": 0.1}'::jsonb
)::integer AS model_id;

Prediction

Copy
SELECT neurondb.predict(
    (SELECT model_id FROM linreg_model),
    features
) AS prediction
FROM test_table;

Learn More

For detailed documentation on regression algorithms, regularization, feature selection, and evaluation metrics, visit:

Regression Documentation

Related Topics

• Random Forest - Random Forest regression
• Gradient Boosting - Gradient boosting regression