Time Series Forecast Reconciliation via Fixed Node Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current approaches for reconciling forecasts from time series data are computationally intensive and require numerous assumptions, failing to efficiently preserve relationships between nodes.
Innovation Solution
A method that generates forecasts by selecting a subset of 'fixed' nodes, which are not reconciled, and performing reconciliation only on 'non-fixed' nodes, using techniques like Lagrange multiplier-based processes to ensure coherence while reducing computational intensity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If reconciliation is performed on all base forecasts, then forecast coherence is improved, but computational complexity increases
Solution Approach 1:
The patent segments the set of base forecasts into two distinct groups: fixed forecasts (which are not reconciled) and non-fixed forecasts (which are reconciled). This segmentation allows the reconciliation process to be applied selectively only to the non-fixed forecasts, thereby reducing computational complexity while still achieving forecast coherence for the reconciled portion. The segmentation principle directly addresses the contradiction by dividing the problem into manageable parts with different processing requirements.
Solution Approach 2:
The patent applies partial action by performing reconciliation on only a subset of forecasts (non-fixed forecasts) rather than all base forecasts. This partial reconciliation approach maintains forecast coherence for the reconciled forecasts while avoiding the excessive computational burden of reconciling every forecast. The method achieves sufficient coherence through partial application of the reconciliation process.
2Reliability
If reconciliation is performed on all base forecasts, then forecast coherence is improved, but computational time increases
Solution Approach 1:
The patent segments the reconciliation process into two stages: fixed forecasts that are excluded from reconciliation and non-fixed forecasts that undergo reconciliation. This segmentation reduces computational time by eliminating unnecessary reconciliation operations on fixed forecasts while maintaining coherence for the non-fixed forecasts that require it. The time savings come from processing only the necessary subset of forecasts.
Solution Approach 2:
The patent implements partial action by applying reconciliation only to non-fixed forecasts rather than all forecasts. This partial application significantly reduces computational time while still achieving the desired forecast coherence for the reconciled forecasts. The method avoids excessive computation by identifying and excluding forecasts that do not require reconciliation.
3Reliability
If reconciliation is performed on all base forecasts, then forecast relationships are preserved, but number of assumptions increases
Solution Approach 1:
The patent segments forecasts into fixed and non-fixed categories, where fixed forecasts are assumed to be reliable and do not require reconciliation. This segmentation reduces the number of assumptions needed because the fixed forecasts are taken as given, eliminating the need to make assumptions about their relationships. Only the non-fixed forecasts require reconciliation assumptions, thereby reducing the overall number of assumptions while preserving necessary forecast relationships.
Solution Approach 2:
The patent applies partial action by making reconciliation assumptions only for non-fixed forecasts rather than all forecasts. This partial application reduces the number of assumptions required while still preserving forecast relationships for the reconciled forecasts. The fixed forecasts are excluded from the assumption-making process, reducing overall complexity.
Data Source
AI summary
A method of generating forecasts from time series data includes receiving a set of time series data organized according to a data structure having a plurality of nodes, generating a plurality of base forecasts, including a base forecast for each node, and selecting a sub-set of the plurality of nodes as fixed nodes. The method also includes performing a reconciliation process to generate reconciled forecasts, where the reconciliation process includes reconciling only the base forecasts of non-fixed nodes, and merging the base forecasts of the fixed nodes and the reconciled forecasts of the non-fixed nodes to generate an overall forecast.