Demand Forecast Disaggregation Using Daily Parameters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Economic modeling in retail environments faces challenges in providing timely and accurate daily demand forecasts, as existing methods are resource-intensive and costly when run on a daily basis, limiting resolution and utility by aggregating data to a weekly scale.
Innovation Solution
A computer-implemented method that transforms transactional data into demand forecasts by estimating daily disaggregating parameters (DDPs) to break down weekly forecasts into daily components, using error minimization techniques and demand models, reducing computational resources and improving forecasting accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the demand model is run on a daily basis to provide daily demand forecasts, then the resolution and utility of the model are improved, but the computational resources and costs increase significantly
Solution Approach 1:
The patent segments the forecasting process into two distinct parts: (1) running the computationally intensive demand model on aggregated weekly data to generate weekly forecasts, and (2) disaggregating the weekly forecasts into daily components using pre-computed daily disaggregating parameters (DDPs). This segmentation allows the system to maintain daily forecasting resolution while avoiding the computational burden of running the full demand model daily.
Solution Approach 2:
The patent performs preliminary computation of daily disaggregating parameters (DDPs) in advance by analyzing historical transactional data to determine the typical proportion of demand that occurs on each day of the week. These pre-computed DDPs are then reused for disaggregating weekly forecasts into daily forecasts, eliminating the need to re-compute these parameters frequently and reducing computational resources required for daily forecasting.
2Use of energy by moving object
If the demand model is aggregated to a weekly basis to reduce computational resources, then the computational costs are reduced, but the resolution and utility of the model are limited
Solution Approach 1:
The patent introduces daily disaggregating parameters (DDPs) as an intermediary mechanism that bridges the gap between weekly aggregated forecasts and daily granular forecasts. The DDPs act as a translator that converts the coarse weekly forecast into fine-grained daily forecasts by applying the pre-computed daily proportions, thereby maintaining high resolution without requiring high computational resources.
3Measurement precision
If the model is run for each day of the week to get desired resolution, then the granularity of the forecast is improved, but the time and resources needed to execute the model increase
Solution Approach 1:
The patent segments the forecasting workload by running the complex demand model only once per week on aggregated data, rather than executing it separately for each day. The daily granularity is then achieved through mathematical disaggregation using pre-computed DDPs, significantly reducing the total model execution time while maintaining the desired level of detail in the forecasts.
Solution Approach 2:
The patent performs preliminary computation of daily disaggregating parameters (DDPs) in advance by analyzing historical transactional data to determine the typical proportion of demand that occurs on each day of the week. These pre-computed DDPs are then reused for disaggregating weekly forecasts into daily forecasts, eliminating the need to re-compute these parameters frequently and reducing computational resources required for daily forecasting.
Data Source
AI summary
A computer-implemented method transforms transactional data into a forecast of demand for controlling a commerce system. Goods move between members of a commerce system. The transactional data is recorded. Daily disaggregating parameters (DDP) are estimated by minimizing an error function of day of week and transactional data grouped according to promotion, price range, or customer. Model parameters are estimated based on the DDP and transactional data using a demand model to generate a weekly forecast of demand for the good. The weekly forecast of demand is disaggregated into daily components using the DDP. The daily components of the forecast of demand are provided to a member of the commerce system to control the movement of goods in the commerce system. Any variation in the demand model due to promotions, price changes, out-of-stock, and low selling product is taken into account when determining the daily components.