Which technique can be used for catering order by multiple stores


#1

Hello friends,

I have a problem statement where multi-channel retailer caters to millions of online orders on a daily basis, and depending on the shipping address, multiple stores can cater to the order.
Which algorithm can be used to identify which store would cater to which order.

Thanks,
Abdul


#2

To me seems an allocation problem, so I would use some linear programming technique.


#3

Thanks for suggestion,

I too thought the same and was looking for Genetic Algorithms to use, but there was nothing to optimize.
The data I had just consist of order number, quantity and stores assigned. I’ve talked to Data Scientist, by looking the data he said it could be done only by hard coding, because the data set I had can’t be learnt by any machine learning technique.

Have a good time:)


#4

Yes, I agree, ML algorithm could not be my first choice in this case.

What about start with an heuristic algorithm, then start to swap in couple links ?

I mean you can just try to allocate just with minimum distance from client to warehouse, then start to swap every link with a near warehouse. You ought to build for every warehouse a list of n (5 ?) warehouse near, and swap the order with every warehouse then see if cost function minimize or not.

First think you must identify your cost function or you will have nothing to optimize.

I followed a MOCC from a crazy Australian man, but I learned a lot and had fun in between.


#5

Yeah, I’ve tried with just some Decision Statements keeping in a loop and it worked.


#6

ML may not be the solution for such problems. It is not a predictive model. It is a prescriptive model.
It is clearly a distribution problem even though you do not have a clear objective function. if you can get data on the location of origin and destination locations, develop a surrogate cost function for transportation and given supply and demand requirements you can come up with a very simple & fast transportation solution. you can adjust the from to cost if there are infeasible shipments between two O-D points.

Srikar


#7

Thanks for suggestion. Have a glance on a data. I just had these three attributes.
OrderNum OrderQty StoresEligible
1 20 S3
2 12 S1
3 19 S1,S2
4 7 S1,S2
5 14 S1,S2,S3