What approach (model) to choose for predicting drug purchases?

Question

Algorithms

I have a task to determine the need for purchases of lek. drugs of a certain type, for a certain region. There is data on purchase requests for the previous five years (which were calculated according to outdated standards). There are several influencing factors both known and unknown, both dynamic and static. Most of the papers that I found on this topic use regression-correlation analysis, but I want to solve this problem using machine learning. Please tell me how this is possible and which of the models or approaches is best suited for solving this problem.
Thanks

Answer 1

Pay attention to machine learning algorithms: scikit-learn.org/stable/tutorial/machine_learning_...
Many algorithms can be used for prediction, you can start with decision trees and random forest ( scikit-learn.org/stable/modules/generated /sklearn.... These algorithms can be used without much customization and do not suffer much from overfitting.Here
is a brute-force tool ( scikit-learn.org/stable/modules/generated/sklearn....
There are also software tools: Weka and Orange
Pandas data manipulation library
NumPy as a base for scientific computing
You can also install Jupyter for clarity If you have any
questions, contact me in the profile.
This is the Python stack of technologies, which, if desired and with due talent, can be mastered on your own.

Answer 2

Actually, it's a service theory problem... There's random consumption, there's random deliveries with random delays, there's storage costs, there are "penalties" for out-of-stock items, items go bad if left unused for some time, etc. .d. etc.
That is, I would start to deal with the subject area, and then I would choose modeling and optimization tools. At least figured out what is the criterion for the success of solving the problem "determining the need for the purchase of medicines ..."

Answer 3

He worked in one of the LARGEST drug supply companies in Russia.
In this company, the owners are related to IT and, according to the old memory, they love IT very much and invest huge sums in the development of IT in their company.
Orders only manually.
Of course, there are some reports, automated tips, etc.
But the decision is made by a person, always adjusting the automation, based on non-formalizable criteria.
In general, the problem is unsolvable.
Or, in other words, it is necessary to leave a person the opportunity to interfere in the work of automation in a way convenient for a person.