Consider engine parts for Aviation & Power Engines. Mfg needs to fill is inventory based on demand & supply.
Now the questions is.. Companies may pile up huge inventory which apparently turns to Slow Moving Inventory/Non Moving Inventory or Dead Stock.
In Operation Research we have Queueing Theory which studys the delay of waiting in line etc.
I would like to check how DataRobot can help in recommending what actions companies can take for Slow Moving Inventory. Eg:- As on Jan'22, my SMI (Slow Moving Inventory) was $100 Million, as a business person, I would like my AI tool to recommend how to reduce my $100 Million month on month.
There are many approaches to consider. One potential approach could be to frame this as a multiclass problem with a range of potential discounting/sales activity to accelerate inventory offloading. This is referred to as Next Best Offer modeling, where the output is are predicted probabilities that sum to 1 for all classes, giving you multiple strategies to pursue with a cascading likelihood of success.
Another strategy could be to maintain sales forecasting models for separate product lines, with variables to account for discounting or marketing activities. You can then use AI Apps to experiment or optimize these activities when sales run lower than expected.
In Queueing Theory information before the start of the queue is generally not considered for modelling and assumptions or actual arrival, departure, dropout, etc rates need to be made or collected to create queue models.
As Mike as has alluded to the different machine learning models that can be applied:
More accurate Time Series Models (TSM) can be created for the abovementioned arrival, dropout, etc. rates to build even better "queue models" or better still a straightforward TSM to predict the different products demand at different times of the year or periods.
This would be more accurate as it also incorporates seasonality and additional features for model assessment into the forecast as opposed to the simple queueing theory assumptions.
Or you could structure it into an OTV (Out-of-Time Validation) model to forecast a product's demand or utilization for the specific month/s or period/s.
Again, in reference to Mike's suggestions there different ways to define your problem/s and solution/s and a better understanding of DataRobot would definitely help you understand your options and the pros and cons of each approach.
The DataRobot University website has different courses and offerings to help you define and achieve your business objectives with the platform: