SAC Smart Predict Hands-On — Classification and Regression
As of 2026-09-25
A hands-on module on SAP Analytics Cloud Smart Predict classification and regression scenarios, sourced directly from SAP's help documentation. Covers the binary-vs-numerical target rule, dataset preparation (acquired vs live, same-source-type and same-HANA-system constraints), the three-section Settings panel (General, Predictive Goal, Influencers), and the Apply Predictive Model workflow: Replicated Column, Apply/Train Date, the ten fixed Assigned Bins, the Outlier Indicator (3x average prediction error), and the Predicted Category / Prediction Probability / Predicted Value / Prediction Explanations outputs. Centres on the single most consequential setting — choosing Prediction Probability with a custom decision threshold over a fixed Predicted Category — and closes with the two-step live-model retraining discipline and what to check before shipping.
What you will learn
- State the exact target-type rule separating classification (binary) from regression (numerical) predictive scenarios in Smart Predict
- Configure the General, Predictive Goal and Influencers sections of the Settings panel for a real classification or regression use case
- Explain the same-source-type and same-HANA-system constraints for training and application datasets, and diagnose a violation
- Choose correctly between Predicted Category and Prediction Probability given a business-defined contact rate, and explain why the Confusion Matrix threshold does not retroactively update an applied output
- Interpret Assigned Bin and Outlier Indicator outputs and use them to decide whether a model is ready to ship or needs retraining
Module overview
Who this is for. You know SAC stories and datasets, and a client has just asked "can SAC predict which customers will churn?" or "can it predict how many complaints we'll get next week?" — questions about a category or a number, not about a time axis. This module covers classification and regression predictive scenarios in Smart Predict, sourced directly from SAP's help documentation; time-series forecasting and its integration with planning models are covered separately in M335. It assumes M333 (AI & LLM Fundamentals) and comfort with datasets in SAC.
Prerequisites
- Working experience with SAC stories and datasets (acquired and live)
- Module M333 (AI & LLM Fundamentals) or equivalent working knowledge
- Access to an SAC tenant with a dataset containing a binary or numerical target column and several candidate influencer columns, for the exercises
Outcomes
- Build and train a classification or regression predictive model end to end, from Settings to a trained report.
- Configure an Apply Predictive Model run that matches a stated business constraint, choosing correctly between Predicted Category and Prediction Probability.
- Diagnose a same-source-type or same-HANA-system violation between training and application datasets.
- Interpret Assigned Bin drift and Outlier Indicator rate to decide whether a model needs retraining.
Full module available to members. The full module adds: the decision framework · the end-to-end scenario walkthrough · the KPI scorecard · the anti-patterns · the code blocks · the knowledge check · the diagrams.