Forecast Reconciliation System Using Segmented Hierarchical Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing forecasting methods struggle to reconcile forecasts across different dimensions of hierarchical data structures, often failing to enforce accounting constraints and scale well with large and complex hierarchies, leading to inconsistencies and inaccuracies.
Innovation Solution
A computer-implemented system and method for performing statistical forecasts that generates forecasts independently for each dimension, then reconciles them using constraints to ensure consistency across dimensions, employing techniques like top-down, bottom-up, and middle-out reconciliation methods, and utilizing optimization routines to minimize loss functions while adhering to aggregation and equality constraints.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If forecasts are generated independently at different levels of hierarchical data, then forecasting flexibility and simplicity are improved, but consistency with accounting constraints deteriorates
Solution Approach 1:
The forecasting process is segmented into independent level-specific forecasting steps followed by a separate reconciliation step. Each level (e.g., product, category, region) can generate forecasts independently using appropriate methods, then the reconciliation algorithm adjusts them to satisfy accounting constraints.
Solution Approach 2:
A reconciliation algorithm acts as an intermediary between independent forecasts at different hierarchical levels. This mediator adjusts the forecasts iteratively to ensure they satisfy accounting constraints while minimizing deviations from original forecasts.
2Reliability
If reconciliation processes are applied to ensure constraint satisfaction, then forecast consistency is improved, but computational complexity increases
Solution Approach 1:
The reconciliation process uses dynamic iterative adjustment where the algorithm adaptively modifies forecasts based on constraint violations. The process continues dynamically until convergence or maximum iterations, adjusting the balance between constraint satisfaction and forecast accuracy.
Solution Approach 2:
The system creates copies of the hierarchical data structure and forecasts to perform reconciliation operations without modifying original data. This allows complex iterative adjustments on copies while preserving the integrity of source forecasts and constraints.
3Productivity
If independent forecasting is performed at each dimension, then forecasting speed is improved, but forecast consistency deteriorates
Solution Approach 1:
The forecasting task is segmented into parallel independent forecasting operations at each hierarchical level, followed by a consolidation reconciliation phase. This segmentation enables speed through parallel processing while maintaining consistency through the subsequent reconciliation step.
Solution Approach 2:
Independent forecasts are generated as preliminary results at each level before the final reconciliation step. These preliminary forecasts capture level-specific patterns and trends, which are then adjusted in the reconciliation phase to ensure overall consistency.
4Quantity of substance
If forecasts are generated for large and complex hierarchies, then comprehensive coverage is improved, but scalability deteriorates
Solution Approach 1:
Large hierarchical forecasting problems are segmented into smaller level-specific subproblems that can be solved independently and in parallel. The reconciliation step then combines these solutions, reducing overall computational burden while maintaining comprehensive coverage.
Solution Approach 2:
The reconciliation process applies partial adjustments only where constraint violations occur rather than reprocessing entire forecast sets. This selective action maintains scalability by focusing computational effort only on necessary adjustments.
Data Source
AI summary
Systems and methods for reconciling a forecast are presented. A method can be used that receives a plurality of hierarchical forecast data sets. An output child data set including an index value and a status indicator representing an unprocessed state is generated. A particular parent data set forecast is identified from a parent data set. Locations for a group of one or more child data set forecasts that are children of the particular parent data set forecast are identified and accessed. A reconciliation operation is performed, a particular child data set forecast is adjusted and stored in a record, and a status indicator for the record is modified.