4 ms·

Yikes, sounds like a pretty tricky mixed integer programming problem. What are the usual go-to techniques for optimization? branch & bound?

by thearn4 12y ago

philip1209 12y ago
We're experimenting with different solving techniques. One of the difficult parts to fit into an algorithm cleanly is "pick how many shifts this employee should have" - e.g. should a part-time employee have two 8-hour + one 4-hour shifts, or one 8-hour + three 4-hour shifts. If you are trying to build that logic into the solver then branch and bound struggles but something like basin hopping seems to work better. However, with fewer employees you can just solve all possible weekly shifts per employee in parallel with a branch and bound. The most basic way to calculate the range is using the hours an employee works in a week and dividing it by both the minimum shift length ("not worth coming in for a shift of less than two hours") and max shift length ("don't have somebody work for more than 8 hours in a day"). Gets more complex when you move into problems like required number of days off per week.
- jaxn 12y ago
  What is your target market? I own several retail stores that usually have 2-6 employees on the clock at any given time and don't see any meaningful benefit at our level. I could see how a place with 20 employees on the clock at one time might have more benefit, but that seems like a pretty small market to me.
  - philip1209 12y ago
    It's a solution searching for a problem. I talked to many coffee shop and restaurant owners, and my conclusion was that independent stores would be hard to acquire as customers, require more ongoing support, and bring in less revenue. Initially I am planning on targeting medical (ERs, private ambulance companies), construction, warehouses, and chains where there are economies of scale or higher-paid hourly workers that would justify such a solution.
    - wlievens 12y ago
      There's a good reason it's called the Nurse Scheduling Problem, not the Barrista Scheduling Problem :-) When my twins were in NICU, it hit me what a challenge it must be to schedule nurses in such a unit. Accross the three shifts in a day, the nurse/patient ratio is a little over one. I think for about 30 NICU babies you had about 75 nurses, so this is an interesting planning problem.

Yikes, sounds like a pretty tricky mixed integer programming problem. What are the usual go-to techniques for optimization? branch & bound?

by thearn4 12y ago

Yikes, sounds like a pretty tricky mixed integer programming problem. What are the usual go-to techniques for optimization? branch & bound?

philip1209 12y ago

We're experimenting with different solving techniques. One of the difficult parts to fit into an algorithm cleanly is "pick how many shifts this employee should have" - e.g. should a part-time employee have two 8-hour + one 4-hour shifts, or one 8-hour + three 4-hour shifts. If you are trying to build that logic into the solver then branch and bound struggles but something like basin hopping seems to work better. However, with fewer employees you can just solve all possible weekly shifts per employee in parallel with a branch and bound. The most basic way to calculate the range is using the hours an employee works in a week and dividing it by both the minimum shift length ("not worth coming in for a shift of less than two hours") and max shift length ("don't have somebody work for more than 8 hours in a day"). Gets more complex when you move into problems like required number of days off per week.

jaxn 12y ago
What is your target market? I own several retail stores that usually have 2-6 employees on the clock at any given time and don't see any meaningful benefit at our level. I could see how a place with 20 employees on the clock at one time might have more benefit, but that seems like a pretty small market to me.
- philip1209 12y ago
  It's a solution searching for a problem. I talked to many coffee shop and restaurant owners, and my conclusion was that independent stores would be hard to acquire as customers, require more ongoing support, and bring in less revenue. Initially I am planning on targeting medical (ERs, private ambulance companies), construction, warehouses, and chains where there are economies of scale or higher-paid hourly workers that would justify such a solution.
  - wlievens 12y ago
    There's a good reason it's called the Nurse Scheduling Problem, not the Barrista Scheduling Problem :-) When my twins were in NICU, it hit me what a challenge it must be to schedule nurses in such a unit. Accross the three shifts in a day, the nurse/patient ratio is a little over one. I think for about 30 NICU babies you had about 75 nurses, so this is an interesting planning problem.