Hierarchical Time Series Forecast Aggregation and Disaggregation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computing resource service providers face challenges in effectively managing and visualizing time series data for forecasting, particularly in aggregating and disaggregating data across multiple hierarchy levels to address supply chain and inventory planning needs.
Innovation Solution
A forecast visualization service that allows users to generate, filter, and override forecasts, enabling data aggregation and disaggregation across dimensions and hierarchy levels, using methods such as proportion-based assignment of override values to lower hierarchy levels, and saving new datasets for visualization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If time series data is aggregated across multiple hierarchy levels for forecasting, then the forecasting coverage and applicability are improved, but the loss of detailed information at lower hierarchy levels increases
Solution Approach 1:
The patent segments the forecasting system into multiple hierarchy levels (e.g., regional, local, facility levels), allowing forecasts to be generated and viewed at different granularities. Each level can be independently analyzed while maintaining connections to parent and child levels, thus preserving detailed information while enabling aggregated forecasting coverage.
Solution Approach 2:
The patent implements a nested hierarchy structure where lower-level forecasts are contained within higher-level forecasts. This allows detailed local forecasts to be nested within regional forecasts, which are in turn nested within national or global forecasts, preserving information at all levels while enabling comprehensive aggregated views.
2Measurement precision
If forecast data is disaggregated to lower hierarchy levels for detailed analysis, then the measurement precision is improved, but the device complexity increases
Solution Approach 1:
The patent creates a universal forecasting system that can operate at multiple hierarchy levels using the same core forecasting engine. The system can generate forecasts at national, regional, local, or facility levels using consistent methodologies, reducing complexity by avoiding the need for separate forecasting systems at each level while maintaining high precision through appropriate data aggregation.
3Adaptability or versatility
If multiple hierarchy levels are maintained for different granularity forecasts, then the adaptability is improved, but the difficulty of detecting and measuring increases
Solution Approach 1:
The patent implements feedback mechanisms that automatically validate and reconcile forecasts across hierarchy levels. When forecasts are generated at different granularities, the system provides feedback to ensure consistency, detect anomalies, and validate data integrity, thereby reducing the difficulty of detecting and measuring across multiple levels while maintaining granularity flexibility.
Data Source
AI summary
In various embodiments described in the present disclosure overrides to forecast data are disaggregated to a lowest hierarchy level and derived forecasts are created based at least in part on the results. In one example, values included in the forecast are organized into a set of dimensions where the dimensions are associated with a hierarchy level of a plurality of hierarchy levels.