随机分配一些元素，但加权到一个列表中，该列表应该只包含整数

Question

我有一个名为employees_summed的len(employees_summed)=9081列表，其中包含员工数量(每个公司)。

我的目标是根据员工的数量，在这个列表中分配一定数量的元素bev_calc，比如bev_calc=2000。

我尝试使用numpy.random.choice，但它没有分配给定的数字，而是返回列表本身的加权平均值：

n = sum(employees_summed)
percentage_list = [x / n for x in employees_summed]
    weightedList = choice(employees_summed, len(employees_summed), percentage_list)

我们的目标是做类似以下的事情：

bev_company = []
percentage = bev_calc / n
for i in employees_summed:
    cars_per_building = percentage * i
    bev_company.append(cars_per_building)

但是由于employees_summed的长度，返回的数字大多小于1，并且round(i)会删除太多的bev_calc，因为它将舍入为0。

有没有办法做到这一点，这样bev_company就有整数作为值，这些整数加起来大致等于2000？随机分布，但根据一家公司有多少员工来权衡？

Answer 1

这不是一种有效的方法，但我认为下面的代码可以满足您的需求：

import numpy as np
np.random.seed(0)

Companies = 10
employees = np.random.randint(5, 100, size = (Companies,))
cars = 100

# make an initial guess of the car distribution and floor its values
guess = employees/employees.sum()*cars
floor_guess = np.floor(guess)

# if the floored values dont add up the the number of cars, scale the 
# distribution such that the company with the closest value is rounded up.
while floor_guess.sum()!=100:
    distances_up = guess%1
    closest = np.argmax(distances_up)
    # if there are more than two companies that are the same distance to 
    # their ceiling, choose one randomly.
    if np.sum(distances_up==distances_up[closest])>1:
        idx = np.random.choice(np.where(distances_up==distances_up[closest])[0])
        guess[idx] = np.ceil(guess[idx])
        floor_guess = np.floor(guess)
        continue
    scale = np.ceil(guess[closest])/guess[closest]
    guess *= scale
    floor_guess = np.floor(guess)