Demand Forecasting | dtlr

Resources

Newsletter

© OBASE 2019. ALL RIGHTS RESERVED

Demand Forecasting

 

The cutting-edge demand forecasting model embodies machine learning algorithms racing on a weekly basis, to be the champion, that will compute the demand for a product in your store for that week. Every Product x Store x Week is appointed with their champion algorithm to calculate the forecast.

 

The demand is forecasted accommodating seasonality, trends, campaigns and calendar effects. Forecasts are built into suggested order quantities and disclosed to the managers in the field, for feedback and inputs.

 

System runs with 10 different Time-Series based algorithms such as;

  Regression Models,

  Exponential Smoothing,

  Holt-Winters,

  ARIMA,

  Mean8Weeks.

 

We incorporated Ensemble Learning techniques in our data-driven demand forecasting model, thus the system combines the strongest aspects of the algorithms into a single method. This increases the accuracy of the forecasts since each algorithm can be sensitive to certain circumstances, while the ensemble approach tests the results of the algorithms, weighs the impact and calibrates the effect.