# Build features for a model

Model / features

generate features or row weights for a model

Page: <https://meridian.online/operations/build-features-for-a-model>

## Status

Runs as a SQL step.

arcform carries it as a `sql:` step today.

## The DuckDB step

features for a model from windows and aggregates

It reads the `orders` table the proofs share. Create it first:

```sql
CREATE TABLE orders AS SELECT * FROM (VALUES
  (1,  'Alice Ng',      'north', DATE '2026-01-03', 120.00, 'widget', 'a|b',   '{"channel":"web","items":[1,2]}'),
  (2,  'alice ng',      'north', DATE '2026-01-05',  80.50, 'gadget', 'b',     '{"channel":"shop","items":[3]}'),
  (3,  'Bob  Okafor',   'south', DATE '2026-01-05', 300.00, 'widget', 'a|c',   '{"channel":"web","items":[4,5,6]}'),
  (4,  'Carla Díaz',    'east',  DATE '2026-01-09',  NULL,  'gizmo',  NULL,    '{"channel":"web","items":[]}'),
  (5,  'Bob Okafor',    'south', DATE '2026-01-12',  45.25, 'gadget', 'c',     '{"channel":"phone","items":[7]}'),
  (6,  'Dev Patel',     'west',  DATE '2026-01-12', 510.00, 'widget', 'a',     '{"channel":"web","items":[8,9]}'),
  (7,  'Carla Diaz',    'east',  DATE '2026-01-20',  99.99, 'gizmo',  'b|c',   '{"channel":"shop","items":[10]}'),
  (8,  'Dev Patel',     'west',  DATE '2026-02-02', 510.00, 'widget', 'a',     '{"channel":"web","items":[8,9]}'),
  (9,  'Eve Larsen',    NULL,    DATE '2026-02-03',  15.00, 'gadget', NULL,    '{"channel":"web","items":[11]}'),
  (10, 'Alice Ng',      'north', DATE '2026-02-10', 220.00, 'gizmo',  'a|b|c', '{"channel":"phone","items":[12,13]}')
) t(order_id, customer, region, order_date, amount, product, tags, meta);
```

Then run the statement:

```sql
CREATE TEMP TABLE result AS
SELECT order_id, customer, amount,
       count(*) OVER w AS prior_orders, coalesce(sum(amount) OVER w, 0) AS prior_spend,
       datediff('day', lag(order_date) OVER (PARTITION BY lower(customer) ORDER BY order_date), order_date) AS days_since_last
FROM orders WINDOW w AS (PARTITION BY lower(customer) ORDER BY order_date ROWS BETWEEN UNBOUNDED PRECEDING AND 1 PRECEDING);
SELECT * FROM result ORDER BY order_id;
SELECT CASE WHEN ((SELECT max(prior_orders) FILTER (WHERE order_id = 10) = 2 AND max(prior_spend) FILTER (WHERE order_id = 10) = 200.5 AND max(days_since_last) FILTER (WHERE order_id = 10) = 36 AND max(prior_orders) FILTER (WHERE order_id = 1) = 0 FROM result)) THEN 'as expected' ELSE error('the earlier orders of a customer were not counted into the later one') END AS verdict;
```

It needs nothing installed beyond DuckDB itself.

## Variants

- Generate features for a model: Derive new feature columns from the table, its related tables or its signals.
- Weight rows for training: Give each row a weight for a model to train on, such as equal total weight per label class.

## What other tools call it

Meridian's classification; an independent blind rating agreed on 86.2% of 312 rows (exact class, version 2).

- Alteryx Designer: [`Build Features`](https://help.alteryx.com/current/en/designer/tools/alteryx-intelligence-suite/machine-learning/build-features.html) (read 2026-09-27)
- mlpack: [`Signal processing`](https://www.mlpack.org/doc/user/preprocessing.html) (read 2026-09-27)
- RapidMiner Studio: [`Generate Weight (LPR)`](https://docs.rapidminer.com/latest/studio/operators/blending/attributes/generation/generate_weight_lpr.html) (read 2026-09-27)
- RapidMiner Studio: [`Generate Weight (Stratification)`](https://docs.rapidminer.com/latest/studio/operators/blending/attributes/generation/generate_weight_stratification.html) (read 2026-09-27)