Course structure
Introduction
- motivation: the large-scale forecasting problem
- layout of the course
Exploring and Processing Timestamped Data
- accumulation: transforming transactional data into time series data
- handling missing and zero-valued intervals
- aggregation: building the data hierarchy
- two feasible data layouts for SAS Forecast Server
- using the TIMEDATA procedure
- identifying systematic variation in the data
- a SAS toolbox for exploring time series data
The Design of SAS Forecast Server-Based Forecasting Software
- the design of SAS Forecast Server-based forecasting software
- functionality overview: system inputs and automatic model selection
Diagnosing and Selecting Models Automatically
- HPFDIAGNOSE procedure: automatic model specification
- model selection lists and automatic model selection
Creating Custom Models and Managing Model Lists
- creating custom models: HPFxxxSPEC procedures
- creating and managing model selection lists: HPFSELECT procedure
Using the Events Functionality
- creating and managing event variables: HPFEVENTS procedure
- using the HPFEVENTS procedure: extensions and further details
Reconciling Statistical Forecasts
- basic forecast reconciliation using the HPFRECONCILE procedure
- disaggregation methods in forecast reconciliation
Producing, Assessing, and Modifying Forecasts
- honest assessment
- preparing to generate forecasts: accumulation and aggregation of the data
- preparing to generate forecasts: creating custom models, diagnosing models, and building a model selection list
- generating forecasts: automatic model selection and forecast outputs
- assessing system accuracy and generating reconciled forecasts
Rolling the Forecasting System Forward in Time
- forecasting by exception
- assessing model degradation
- handling data updates and structural changes
