Inventory Demand Forecasting Using Seasonal Pattern Segmentation

The patent segments historical demand data into multiple seasonal patterns (e.g., annual, semi-annual, quarterly, monthly seasonality) and applies separate Fourier series models to each pattern. This allows the system to capture both regular periodic seasonality and irregular seasonal variations by combining multiple segmented models rather than relying on a single periodic model.

The patent introduces dynamic adjustment mechanisms where the Fourier series models can adapt to changing demand patterns over time. The system dynamically updates seasonal parameters and can switch between different seasonal patterns based on observed demand behavior, enabling accurate forecasting even when seasonal patterns become irregular or non-periodic.

The patent implements feedback mechanisms where actual demand data continuously updates and refines the Fourier series demand models. The system compares forecasted demand with actual demand, identifies deviations and seasonal patterns, and adjusts future forecasts accordingly. This feedback loop enables more accurate demand prediction, reducing both stockouts and excess inventory accumulation.

The patent dynamically changes inventory parameters (such as safety stock levels, reorder points, and order quantities) based on the identified seasonal patterns and forecast accuracy. During high-seasonality periods, the system adjusts parameters to ensure availability, while during irregular or low-demand periods, it reduces parameters to prevent excess inventory, optimizing the balance between availability and inventory quantity.