Unified Data Mart for Retail Forecasting Accuracy and Model Rebuilding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing forecasting systems require maintaining two separate data marts for accuracy monitoring and model rebuilding, which is resource-intensive and logistically challenging, especially for retailers with large datasets, as it necessitates significant time and space investments.
Innovation Solution
A system that allows concurrent operation of forecast accuracy monitoring and model construction using production data, enabling realignment and reuse of data to assess and rebuild models without interfering with the production environment, thereby reducing the need for separate data marts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If two separate data marts are maintained for accuracy monitoring and model rebuilding, then forecast accuracy can be monitored and models can be rebuilt, but resource consumption and operational complexity increase significantly
Solution Approach 1:
The patent merges the accuracy monitoring data mart and model rebuilding data mart into a single shared data mart structure. Both the forecast accuracy monitoring system and the model construction system access and operate on the same production data, eliminating the need for separate data copies while maintaining system reliability and operational independence through logical separation of functions.
2Ease of operation
If two separate data marts are maintained for accuracy monitoring and model rebuilding, then independent operation is enabled, but time and space resources are significantly consumed
Solution Approach 1:
The patent combines data storage resources while maintaining operational independence through logical system separation. The single shared data mart allows both accuracy monitoring and model rebuilding operations to access the same data simultaneously without the time delays associated with copying and synchronizing separate data marts, thus reducing model rebuilding time while preserving independent operation capabilities.
3Reliability
If two separate data marts are maintained for accuracy monitoring and model rebuilding, then data isolation is achieved, but space resources are significantly consumed
Solution Approach 1:
The patent merges multiple data mart requirements into a single shared data mart that serves both accuracy monitoring and model rebuilding functions. This eliminates the need to store duplicate copies of production data in separate data marts, significantly reducing space consumption while maintaining production environment stability through controlled access and operational independence.
4Ease of operation
If separate data marts are used for accuracy monitoring, then operational independence is maintained, but logistical problems arise during updates and maintenance
Solution Approach 1:
The patent consolidates data mart maintenance operations into a single shared data structure, eliminating the need to synchronize updates across multiple separate data marts. Both the accuracy monitoring system and model construction system access the same updated data simultaneously, simplifying maintenance operations while preserving system independence through logical separation of access patterns and operational workflows.
Data Source
AI summary
Computer-implemented systems and methods are provided to perform accuracy analysis with respect to forecasting models, wherein the forecasting models provide predictions based upon a pool of production data. As an example, a forecast accuracy monitoring system is provided to monitor the accuracy of the forecasting models over time based upon the pool of production data. A forecast model construction system builds and rebuilds the forecasting models based upon the pool of production data.