.. _auto_ml_predict_ensemble:

*ML PREDICT (ENSEMBLE)*
=======================

Runs a list of trained models (e.g. one per batch from a memory-bounded batch-ensemble run) over a dataset and AVERAGES their per-target predictions (bagging). When the dataset also carries the actual target columns, reports held-out R^2 of the averaged prediction. The predict/eval end of the batch-ensemble ML pipeline.

When to use
-----------

Tagged: ``average``, ``bagging``, ``batch``, ``ensemble``, ``ml``, ``predict``, ``r2``.

Inputs
------

.. list-table::
   :header-rows: 1
   :widths: 20 20 20 20 20 20

   * - Label
     - ID
     - Type
     - Default
     - Required
     - Description
   * - Models
     - models
     - text
     - —
     - ✓
     - Comma-separated list of trained model .pkl paths (one per batch). Also accepts a dataset/list of model-path rows.
   * - Dataset
     - dataset
     - dataset
     - —
     - ✓
     - Rows to predict (already cleaned/normalized/PCA-transformed the same way the models were trained). If the rows also contain the actual target columns, held-out R^2 is computed.
   * - Target Columns
     - targets
     - text
     - —
     - 
     - Comma-separated target column names to score (e.g. SOC_p9). Defaults to every predicted '<target>_pred' column with its suffix stripped.

Outputs
-------

.. list-table::
   :header-rows: 1
   :widths: 20 20 20 20

   * - Label
     - ID
     - Type
     - Description
   * - Ensembled Predictions
     - predictions
     - dataset
     - One row per input row with the averaged '<target>_pred' values (plus an id).
   * - Held-out R^2
     - r2
     - string
     - Pooled R^2 of the averaged prediction vs the actual targets (null when no actuals present).
   * - Per-target R^2
     - per_target_r2
     - dataset
     - Per-target R^2 rows ({ target, r2, n }).
   * - Models Used
     - models_used
     - integer
     - How many of the supplied models produced usable predictions.
   * - Status
     - status
     - string
     - Human-readable summary.

Disciplines
-----------

- ai_ml.prediction

.. raw:: html

   <hr style="margin-top:2em">
   <p style="font-size:11px;color:#888">
   Auto-generated from <code>platform</code> schema. Worker id: <code>ml_predict_ensemble</code>. Schema hash: <code>f6b50cdf40b7</code>. Hand-curated docs in <code>workerexamples/</code> override this page when present.
   </p>