This paper explains how the Demand Forecasting Machine Learning (ML) solution helps different business teams provide forecasts several Demand Signals such as: 

• New contracts and shipments forecast: Total contracts that are active during reporting fiscal week.
• Active service units (ASUs) forecast: Total unique assets that have qualifying warranties that are active during the reporting fiscal week time. 
• Service requests (SRs) forecast: Cases raised or opened during the reporting fiscal week time period.
• Dispatches and tasks forecast: Parts sent based on open cases during the reporting fiscal week time period segments.

Demand Forecasting enables business teams to better understand the upcoming demands and consumption based on historical trends, patterns, and seasonality behaviors of demand and use forecast to better prepare and develop their financial plans such as Annual Operating Plan (AOP) and budget planning for the upcoming fiscal year. Also, users in operational planning roles use these forecasts to develop call center head count plans, parts procurement plans, and other plans in a more accurate and effective manner.

This paper explains additional improvement models that include classification and clustering approaches to identify the best timeseries to use based on the accuracy of the forecasts. This document also discusses how low and high volumes are handled and how the impact of COVID-19 and holidays in forecast predictions is measured.