- Joined
- Jul 17, 2016
- Messages
- 2,001
No. The platform I use (Snowflake) doesn’t support R, I think. Python is supported, but I’m not familiar with it, yet. The data don’t have the dependent variable, so logistic regression is not possible.Just out of curiosity .. did you made a regression model in R?
They’re looking for a cheap way to detect phantom inventory, because getting it from a third party is too expensive. The contract to that vendor has not been renewed. I bet somebody thought that they can solve it cheaply by throwing maybe $10K to this Indian consulting company. We know how that went.
Phantom inventory detection is anything but a new problem. The company’s commercial analytics team brags that they use AI with 200+ parameters to create a fairly accurate demand forecast. The thing is, our team is low on the totem pole and doesn’t have access to it. That’s corporate bureaucracy to you.
In any case I’m not a programmer nor a statistician, I’m a corporate manager well outside the IT division who doesn’t even have admin access. I left IT about a decade ago, and I prefer not to get dragged back into it. My competence is in SQL and VBA, but I don’t advertise it because I get pigeonholed as one of the nerds, which is detrimental to my career. I’m slowly learning Python, but I’m not advertising it either.
I created the model entirely in SQL using only functions available in Snowflake. The consultant guy supposedly had access to an offshore support team having Python programmers, but since he did nothing then the team wasn’t engaged. I guess it’s a silver lining of some sort, because SQL is a data manipulation language, so it’s more efficient when churning huge amount of data (200 million records).
If you’ve done Six Sigma you should be familiar with the concept. Find the data points outside the control limits. The data in their original form work poorly with average and standard deviation, just too many zeros. I had to be creative and transform them into strings to be usable.
