I have a sales data for few years, I want to distribute data for few quarters. I can simply divide a that sales number over four quarters, but I would like to have it like a gaussian distribution or sinusoidal distribution kind of using python code. Any idea would be of great help. Asking for the first time in stack overflow, forgive me if I there is some mistake in my presentation.
| Year | Sales |
|---|---|
| 2020 | 1400 |
| 2021 | 1800 |
| Quarter | Sales |
|---|---|
| 2020Q1 | 300 |
| 2020Q2 | 500 |
| 2020Q3 | 400 |
| 2020Q4 | 200 |
when I sum up all quarter values it sums up to 1400 as above.
I tried to check if there are few modules, I found there is a module called sympy but dont know how to fit my problem in that.
I don't think you need SymPy for this. The python random module has the ability to generate random numbers in a Gauss distribution but this is going to be hard to see on only 4 bins. One approach would be to generate random numbers (with some distribution), scale them to get close to the answer, add any deficit, and then sort them and return them centered on the maximum value to get some peak symmetry:
import random
def g(bins, total, mu, si):
pts = [abs(random.gauss(mu, si)) for i in range(bins)]
mag = total/sum(pts)
pts = [int(i*mag) for i in pts]
defic = total - sum(pts)
while defic:
random.shuffle(pts) # distribute deficit randomly
pts[0] += 1
defic -= 1
pts = sorted(pts)
rv = []
while pts:
if len(rv)%2:
rv.append(pts.pop())
else:
rv.insert(0, pts.pop())
return rv
e.g.
>>> g(10,1000,1000,100)
[91, 96, 103, 107, 112, 109, 105, 98, 94, 85]
>>> g(10,1000,100,100)
[9, 47, 63, 154, 360, 161, 125, 49, 30, 2]
>>> sum(_)
1000