SAP Analytics Cloud Predictive Analytics (Chabert, AntoineSerre, David)（Z-Library）

# SAP Analytics Cloud Predictive Analytics — Reading Guide ## 【One-Line Pitch】 A practical, end-to-end guide for building, training, and applying predictive models in SAP Analytics Cloud—covering time series forecasting, classification, and regression—written for financial planners, data analysts, and data scientists who want to turn historical data into forward-looking business decisions. --- ## 【Book Arc】 - **Opening (~0%–9%)**: Introduces predictive analytics as a branch of data analytics using historical data and statistical/machine learning methods to forecast future outcomes. Sets up the book's structure: three parts covering fundamentals, time series forecasting, and classification/regression, with a target audience spanning financial planners, data analysts, and data scientists. - **Early (~9%–28%)**: Explains the core object—the predictive scenario—and walks through predictive analytics project planning, including stakeholder roles (business owners, data engineers, prediction creators, story designers, prediction consumers) and an implementation methodology. Emphasizes that no advanced math or data science skills are required to operate SAP Analytics Cloud's predictive features. - **Early–Middle (~28%–38%)**: Dives into time series forecasting fundamentals, focusing on data sources: planning models (predictive planning integration) and datasets. Covers technical compatibility rules between training and prediction data sources, including acquired vs. live dataset requirements. - **Middle (~38%–53%)**: Details dataset preparation for time series forecasting: minimum structure (date + quantity columns), optional influencer columns (external regressors), and entity columns for sliced predictions (up to five entity columns, different training dates per entity). Includes concrete examples like cash flow and sales forecasting. - **Late (~53% onward)**: Moves into hands-on workflows: creating time series forecasting predictive scenarios, integrating forecasts into enterprise planning processes, automating forecast delivery via multi-actions, and best practices for improving model accuracy. The book concludes with explanations of the automated "smart predict" techniques powering these models. --- ## 【Key Takeaways】 - **Predictive scenarios are the central object** (Early): SAP Analytics Cloud uses three scenario types—time series forecasting, classification, and regression—as the foundation for creating and delivering predictions. Understanding which scenario fits your business question is the first design decision. - **Predictive planning bridges analytics and planning** (Early): Only time series forecasting integrates with enterprise planning models, enabling forecasts to run directly within planning contexts. This is the key differentiator for finance teams using SAP Analytics Cloud for budgeting and forecasting cycles. - **Stakeholder clarity drives project success** (Early): A predictive project involves distinct roles—business owners (sponsors), data engineers (data preparation and pipelines), prediction creators (model builders), story designers (visualization), and prediction consumers (decision-makers). One person can fill multiple roles, but each function must be covered. - **Data source compatibility is enforced** (Middle): When using datasets, training and prediction datasets must match in type (both acquired or both live), and live datasets must reside in the same SAP HANA instance. This constraint shapes how you architect your data flows. - **Dataset structure follows simple rules** (Middle): For time series forecasting, a training dataset needs at minimum a date column (recognized date format, not integers) and a numeric quantity column. Each row is one observation, ordered by date. - **Influencers improve forecast accuracy** (Middle): Optional influencer columns (external regressors) can be numeric or nominal, but you must provide values for the entire forecasting horizon—otherwise smart predict assumes defaults, potentially producing unrealistic predictions. - **Entities enable granular predictions** (Middle): Entity columns let you generate separate forecasts for different slices (e.g., country, product). Smart predict supports up to five entity columns and handles entities with different training date ranges, accommodating missing data or new entities. --- ## 【Reading Tips】 - **Skim Part I if you're experienced**: Chapters 1–3 cover predictive analytics basics and project methodology. If you already know SAP Analytics Cloud fundamentals, jump to Chapter 4 for data preparation rules—this is where practical work begins. - **Deep-read the dataset structure sections**: The tables showing training dataset layouts (date, quantity, influencers, entities) are the most reusable reference material. Bookmark these for when you build your own datasets. - **Pay attention to the planning model vs. dataset distinction**: These two data source types behave differently—planning models read structure transparently, while datasets require explicit formatting. Knowing which path you're on saves significant debugging time. - **Don't skip the stakeholder chapter**: Even if you're a solo practitioner, the role definitions (especially prediction creator vs. data engineer) clarify what tasks you need to handle yourself and what you might delegate. - **Use the worked examples as templates**: The cash flow and sales forecasting examples (with influencer and entity columns) are directly adaptable to your own business data—treat them as starting templates rather than just illustrations. --- ## 【Coverage Limits】 This guide covers the book's opening through the middle sections (approximately 53% of the book), focusing on predictive analytics fundamentals, project methodology, and time series forecasting data preparation. The later chapters on step-by-step scenario creation, planning integration, automation via multi-actions, and the internal "smart predict" algorithms are not covered in detail here. --- ##