Hello,
Time series forecasting is typically used to forecast the evolution of a given KPI per time, not really specific dates.
Examples:
- Evolution of my sales per store
- Evolution of my expenses per cost center & cost type
In your example, you are rather trying to predict the likeliness of a certain event - do I need to refill the container at this date?
My suggestion is to consider classification for doing so and predict the likeliness of having to refill the container for the specific date. As always the beauty as well as the difficulty of data science lies in the data preparation phase where you typically need to address & solve 80% of the problem. Obvious challenges to resolve in your use case is how to formulate dates - not absolute but relative to past events, months etc. For instance 90 days have passed from the last refill to today etc.
In addition to the dates handling, are there any factors (features) that we can relate to the event "the container was refilled"?
A similar idea / example can be found here https://towardsdatascience.com/forecasting-of-periodic-events-with-ml-5081db493c46
Best regards,
Antoine (product manager for SAP Analytics Cloud predictive scenarios)
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.