Demand Forecast Combination System for Supply Chain Planning

The patent segments the demand forecasting process into multiple independent data streams with different granularities and time horizons. Instead of relying on a single coarse-grained demand planning forecast, the system divides the forecasting function across multiple sources including demand planning (long-term, low granularity) and responsive replenishment (short-term, high granularity), allowing each stream to contribute its strengths to the overall forecast accuracy across different time horizons.

The patent adds a new dimension to demand forecasting by integrating data streams with different temporal resolutions and granularities. By combining forecasts from multiple independent sources that operate at different time intervals (weekly/monthly vs. daily), the system creates a multi-dimensional forecasting approach that simultaneously addresses both long-term planning needs and short-term accuracy requirements.

The patent segments the forecasting responsibility by time horizon, assigning short-term accuracy-critical periods to responsive replenishment data streams while relying on demand planning streams for long-term horizon coverage. This segmentation allows each data stream to operate within its optimal accuracy range, with the combination providing both short-term precision and long-term extendability.

The patent changes the temporal parameters of forecast combination dynamically, using responsive replenishment data for near-term periods where its daily granularity provides high accuracy, and transitioning to demand planning data for longer periods where its coarser granularity is acceptable. The system adjusts the weight and time horizon parameters of each data stream based on the specific forecasting needs.

The patent merges multiple independent demand forecast data streams into a single combined forecast that leverages the strengths of each source. By combining demand planning data (providing long-term horizon coverage) with responsive replenishment data (providing high granularity and short-term accuracy), the system achieves both extended planning capability and improved demand accuracy that neither stream could provide alone.

The patent creates a composite forecasting approach by integrating multiple data streams with different characteristics (granularity, time horizon, accuracy profiles) into a unified forecast. This composite forecasting structure allows the system to select or weight different data sources based on the specific forecasting requirements, achieving optimal balance between time horizon and accuracy.