Concept card

SAP HANA Automated Predictive Library (APL)

As of 2026-09-25

Edited by Cédric-Olivier Mâry, sole editor and publication director — 27 years in the SAP ecosystem. How these pages are produced, and where automation is used.

What is SAP HANA Automated Predictive Library (APL)?

APL is SAP HANA's automated-ML AFL component: SAP's own sample code shows the APL/VariableAutoSelection flag doing the feature-selection work a PAL call leaves to the caller, across a function set spanning classification, regression, forecasting, key influencers, recommenders, social-network modeling and data-drift monitoring. It's the one library SAP names directly as what Smart Predict delegates to on live on-premise HANA connections.

What APL automates, and how SAP's own code proves it

The Automated Predictive Library (APL) is SAP HANA's automated-machine-learning counterpart to PAL (C343): both are AFL — Application Function Library — components, both compute inside HANA, but APL trims the surface a caller has to configure. Where a PAL classification call passes an explicit parameter table naming an algorithm and its hyperparameters (C343), SAP's own APL sample code for a classification model sets three configuration entries and lets APL do the rest: APL/ModelType ("binary classification"), APL/CuttingStrategy ("random with no test"), and — the one that matters most — APL/VariableAutoSelection set to true. That single flag is the concrete, documented mechanism behind the word "automated" in APL's name: the library decides which of the input variables actually belong in the model, rather than a data scientist deciding it upfront.

APL runs inside its own schema, SAP_PA_APL, with dedicated table types for models, logs, summaries and debrief metrics (MODEL_BIN_OID, OPERATION_LOG, SUMMARY, DEBRIEF_METRIC_OID, DEBRIEF_PROPERTY_OID) — the schema name alone is worth recognizing in a client's system as evidence APL is installed and in use.

Why it matters

APL/VariableAutoSelection is the concrete, documented mechanism behind "automated" in APL's name — a consultant can point to this specific flag instead of describing automation vaguely.
APL's function set is materially broader than the three scenarios Smart Predict exposes in the SAC UI (C341): recommenders, social-network modeling and data-drift monitoring are APL capabilities without a direct Smart Predict UI equivalent.
SAP names APL directly as what Smart Predict delegates to for live on-premise HANA connections — the one place in the documentation that ties C341, C343 and C344 together concretely.

Key points

APL is HANA's automated-ML AFL component; SAP's own sample code sets APL/ModelType, APL/CuttingStrategy and APL/VariableAutoSelection=true instead of naming an algorithm and tuning parameters.
APL runs in its own SAP_PA_APL schema with dedicated table types for models, logs, summaries and debrief metrics.
Predictive model functions (lifecycle): CREATE_MODEL(_AND_TRAIN), TRAIN_MODEL, RETRAIN_MODEL, UPDATE_MODEL, APPLY_MODEL(_AND_TEST), TEST_MODEL, PUBLISH_MODEL, EXPORT_APPLY_CODE, GET_MODEL_INFO/DEBRIEF, GET_VARIABLE_INDICATORS.
Predictive business functions (scenario coverage): FORECAST/FORECAST_AND_DEBRIEF (time series), KEY_INFLUENCERS, RECOMMEND family (recommenders), CREATE_SOCIAL_MODEL_AND_TRAIN/APPLY_SOCIAL_MODEL (social-network modeling), PROFILE_DATA family, SCORING_EQUATION.
Data drift detection is a distinct, documented APL capability with learn/report and detect/report examples, including segmentation by a dimension like country.
EXPORT_APPLY_CODE lets APL generate scoring code deployable outside HANA — but the automated variable-selection decisions behind it remain APL's.
Called via direct SQL against AFL procedures, or via hana-ml's hana_ml.algorithms.apl module in Python — both paths run the same automated logic.
SAP's own Smart Predict live-connectivity documentation states execution is delegated to APL for on-premise live connections — the one place this dependency is named directly.

Terms used on this page

APL/VariableAutoSelection: The APL configuration flag documented in SAP's own sample code that turns on automated feature/variable selection.
SAP_PA_APL: The dedicated HANA schema APL runs in, with its own model, log, summary and debrief table types.
EXPORT_APPLY_CODE: APL function that generates scoring code deployable outside SAP HANA, based on a trained APL model.
Data drift (APL): APL's dedicated capability to learn a baseline distribution and detect and report deviation from it, optionally segmented by a dimension.
SCORING_EQUATION: APL function that exports a trained model as a portable scoring formula.

Sources

Full card available to members. What the full card adds: the full decision framework · the common pitfalls and their fix · the cheat sheet · the code blocks · the facts worth quoting.

Open in the app →