This course teaches students to develop and maintain a large-scale forecasting project using SAS Visual Forecasting tools. For the course project, students build and then refine a large-scale forecasting system. Emphasis is initially on selecting appropriate methods for data creation and variable transformations, model generation, and model selection. Students are then asked to improve overall baseline forecasting performance by modifying default processes in the system.

Learn how to

• Convert transaction sequences to time series data.
• Accommodate a hierarchical structure associated with time series to be forecast.
• Choose appropriate methods for handling missing values in time series data.
• Understand global systematic variation in time series that drives choices for generated model families(for example, ARIMAX, exponential smoothing).
• Understand the fundamentals and functionality restrictions embedded in model families.
• Select useful subsets of exogenous (independent) variables for input variable selection. (Note: This directly builds on the first course.)
• Understand the impact of calendar-based events and other events on forecasting performance.
• Create event variables that improve the precision of forecasts.
• Create partitions of project time series data to apply honest assessment in champion model selection.
• Interpret static effects in fitted models.
• Reconcile generated forecasts.
• Understand best practices associated with performing forecast overrides.
• Efficiently accommodate data updates and manage the forecasting system.

Who should attend

Analysts interested in augmenting their machine learning skills with analysis tools that are appropriate for assaying, modifying, modeling, forecasting, and managing data that consist of variables that are collected over time

Formats available	Standard Duration
Classroom:	3 days
Live Web Classroom:	6 half-day session(s)	System Requirements
e-Learning:	21 hours/180 day license	System Requirements

The courses is primarily syntax based, so analysts taking this course need some familiarity with coding. Experience with an object-oriented language is helpful, as is familiarity with manipulating large tables.

This course addresses SAS Viya, SAS Visual Forecasting software.

Introduction to Large-Scale Forecasting

• The large-scale forecasting problem.
• 1.2 forecasting system overview.

Exploring and Processing Timestamped Data

• Accumulation: Transforming transactional data into time series data.
• Handling missing and zero-valued intervals.
• Aggregation: Building the data hierarchy.
• Packages for TSMODEL.

Automatic Forecasting: Model Specification and Selection

• Introduction to automatic forecasting.
• Automatic model specification.
• Automatic forecast generation.

Creating Custom Models and Managing Model Lists

• Custom models in the forecasting system.
• The Time Series Model (TSM) package.
• Adding custom models to the automatic forecasting system.

Events Variables in the Forecasting System

• Introduction to event variables.
• Creating event variables in the ATSM package.
• Creating event variables with the HPFEVENTS procedure.
• BY-group processing for event variables.

Reconciling Statistical Forecasts

• Introduction to reconciliation.
• Basic forecast reconciliation.
• Disaggregation methods.
• Bottom-up reconciliation.

Setting Up the Forecasting System and Generating Best Forecasts

• Honest assessment and baseline performance.
• Combined model forecasts.
• Outlier detection.
• Conditional processing and error catching.
• Rolling the forecasting system forward in time.