ARIMA forecasting technique for MT cloud

The Auto Regressive Integrated Moving Average (ARIMA) algorithm is added to the forecast engine.

This feature uses a new SCP Artificial Intelligence (AI) service. The SCP AI service is developed as a REST service that provides end points to allow the calling of various Machine Learning/AI packages implemented in Python. Python is widely used in the Machine Learning forecasting community. This enhancement implements ’Auto ARIMA’ in python using pmdarima.

Auto ARIMA Implements SARIMA: Seasonal Auto Regressive Integrated Moving Average. ARIMA and SARIMA are techniques in time-series forecasting and are described in many places.

Auto ARIMA uses various tests to determine the level of differencing required and fits multiple ARIMA models with various different values for the auto-regressive and moving average parameters for the non-seasonal and seasonal elements of the model. The model with the lowest Information Criterion is selected.

The ARIMA algorithm is not included in the system ‘Best Fit’ process.

The AI service currently supports single forecasting calls, whereby the forecast input data for a single forecast entity (item-location combination) is passed to the server and the engine result is returned. This is not optimized for batch processing of large data volumes, many item-location combinations. ARIMA is recommended as an alternative ’Given’ algorithm on an adhoc basis for individual combinations from the worksheet using the ’Fit forecast with options’ feature, or for specific cases where Best Fit or other default forecast engines are not producing desired results.

These forecast engines are added:

  • ARIMA CMonths
  • ARIMA Months
  • ARIMA Weeks

A new measure and calculation is added indicating the count of item-location combinations in the current context with a forecast calculated based on the ARIMA engine:

  • Forecast Algorithm ARIMA (DPLS_FALGORITHM_11)

This worksheet is updated:

  • Forecast Information
Note: This feature is enabled on update for MT Cloud deployments only, however related content changes for ARIMA are only available after loading the demand planning template (dpls.zip) for this version. You are not required a new role or privilege access to use this feature.