Step 7. Sales and demand forecasting: Trainer-> Training

Trainer – Training – Inventory Demand Forecasting

Use Training screen to train prediction models. Users don’t need to do any adjustments to the configuration as system is automatically performing training of your data with predefined parameters with all algorithms for all Product->Customer->Location combination and it is automatically selecting for user the most accurate and most fitting algorithm and prediction method. Predictions results of all Product->Customer->Location are then aggregated in the forecaster and the system will provide Product->Location sales predictions.

Note: Machine learning is an intense CPU process and can occupy 100% of the CPU while performing training of algorithms. Therefore, the right hardware configuration is the key and should be aligned with the dataset size to be processed on a single machine as this will affect the duration and length of the algorithm training.

Model Training

The main screen used to train algorithms. It is as simple as pressing the button. User will see on the screen progress of algorithm training Duration of training is strongly dependent on the hardware used and also on data volume and can take anything from seconds to several hours.

Model Testing

This screen is not used by regular users and it is used for validation of individual algorithms in case of issues with forecasting.