Print Demand Forecasting Using Variable Sampling Intervals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing demand forecasting methods for print production environments face difficulties in accurately forecasting demand components with high variability, particularly when using the same sampling interval throughout disaggregation, forecasting, and aggregation, which can lead to challenging forecasting and convergence issues.
Innovation Solution
A print demand forecasting system and method that processes aggregated demand data to separate demand components into low and high variability sets, allowing for different time scales and using appropriate forecasting techniques for each, with a computer-implemented service manager to adjust and re-aggregate the data for improved forecasting accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the same sampling interval is used throughout disaggregation, forecasting, and aggregation, then the forecasting process is simplified, but forecasting accuracy deteriorates due to difficulty in forecasting demand components with high variability
Solution Approach 1:
The patent segments the forecasting process into distinct phases: disaggregation phase with one sampling interval, forecasting phase with potentially different sampling intervals, and aggregation phase. This allows each phase to use the most appropriate sampling interval for its specific requirements, resolving the contradiction between process simplicity and forecasting accuracy.
Solution Approach 2:
The patent introduces dynamic sampling intervals that can change across different forecasting tasks and demand components. The system adapts the sampling interval based on the variability characteristics of each demand component, using finer intervals for high-variability components and coarser intervals for low-variability components, thereby improving accuracy without uniformly complicating the entire process.
2Measurement precision
If demand data is disaggregated into multiple demand components, then forecasting accuracy for individual components improves, but the overall forecasting system complexity increases
Solution Approach 1:
The patent segments demand data into multiple demand components based on their variability characteristics (e.g., cyclic vs. non-cyclic patterns). Each segment is then processed with appropriate forecasting methods, allowing accurate forecasting of individual components while managing overall system complexity through modular processing.
Solution Approach 2:
The patent changes the parameters of the forecasting system by allowing different sampling intervals and forecasting methods for different demand components. This parameter differentiation enables accurate forecasting of diverse demand patterns without requiring a single complex unified model.
3Measurement precision
If different time scales are used for demand components, then forecasting accuracy improves, but data correspondence and aggregation become more difficult
Solution Approach 1:
The patent introduces an additional dimension of time scale correspondence, mapping demand components from their native time scales to a common aggregation time scale. This dimensional transformation allows accurate representation of demand at multiple time scales while enabling straightforward aggregation through systematic correspondence rules.
Data Source
AI summary
A system and method are provided for performing forecasting with respect to total demand data (including two or more demand components) collected in a print production environment. Each demand component (comprising a set of demand component related points corresponding with a first time scale) may be processed in such a way that a demand component related point can be forecasted with demand component related points corresponded with the second time scale. Both the forecasted demand component related point and demand component related points corresponded with the second time scale can then be corresponded with the first time scale.