Inventory Simulation - for test data

June 5, 2023

[1]: import pandas as pd

import numpy as np

[2]: product_details = pd.read_csv('C:/Users/nissy/Desktop/MTECH/M Project/Code/

,→ZKeras Model/product_data_for_simulation.csv')

[3]: product_details.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 400 entries, 0 to 399
Data columns (total 14 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 index 400 non-null int64
1 id 400 non-null int64
2 unit cost 400 non-null int64
3 lead time 400 non-null int64
4 demand probability 400 non-null float64
5 initial inventory 400 non-null int64
6 mean demand 400 non-null int64
7 sd 400 non-null int64
8 demand distribution 400 non-null int64
9 r_opt 400 non-null float64
10 Q_opt 400 non-null float64
11 Min_Cost 400 non-null float64
12 r_pred 400 non-null float64
13 Q_pred 400 non-null float64
dtypes: float64(6), int64(8)
memory usage: 43.9 KB

[4]: class Product:

def __init__(self, i):
"""
:type i: int - Product number
"""
self.i = i

self.initial_inv = product_details['initial inventory'].iloc[i - 1]

1
 self.mean = product_details['mean demand'].iloc[i - 1]
self.sd = product_details['sd'].iloc[i - 1]

self.probability = product_details['demand probability'].iloc[i - 1]

self.lead_time = product_details['lead time'].iloc[i - 1]

self.demand_lead = self.lead_time*self.probability*self.mean

self.unit_cost = product_details['unit cost'].iloc[i - 1]

self.pred_r = product_details['r_pred'].iloc[i - 1]
self.pred_Q = product_details['Q_pred'].iloc[i - 1]

self.opt_r = product_details['r_opt'].iloc[i - 1]
self.opt_Q = product_details['Q_opt'].iloc[i - 1]

self.demanddist = product_details['demand distribution'].iloc[i - 1]

#self.selling_price = product_details['selling price'].iloc[i - 1]

self.holding_cost = self.unit_cost*0.02
self.ordering_cost = 1000

[5]: def daily_demand(mean, sd, probability, demanddist):

random_num = np.random.uniform(0, 1)
if random_num > probability:
return 0
else:
if demanddist ==1:
y = round(np.random.normal(mean, sd), 0)
elif demanddist == 2:
y = np.random.poisson(mean,1)[0]
elif demanddist == 3:
y = np.random.negative_binomial(n=mean**2 / (sd**2 - mean), p=mean/␣
,→(sd**2), size=1)[0]

else:
#scaling down teh lognormal generated demand
y= round(np.random.lognormal(mean=np.log(mean), sigma=sd,␣
,→size=1)[0]*0.2, 0)

if y > 300 and y < 500:

y= y*0.5
elif y > 500 and y < 1000:
y = y*0.3
else:
y =144

2
 return y

[6]: def cc_monte_carlo_ray(product, q, r):

inventory = product.initial_inv
mean = product.mean
sd = product.sd
lead_time = product.lead_time
probability = product.probability

dist = product.demanddist

order_placed = False
order_time = 0
stock_out = 0

data = {'day':[],'initial inv level':[], 'daily demand': [],'units sold':␣

,→[],

'current inv level': [], 'units lost': [],

'orders': [], 'stockouts':[]}

for day in range(1, 365):

data['day'].append(day)
if day == 1:
day_demand = 0
else:
day_demand = daily_demand(mean, sd, probability, dist)

data['daily demand'].append(day_demand)

data['initial inv level'].append(inventory)

if inventory <= r and not order_placed:

# Time to place an order
order_placed = True
order_time = day

if order_placed and (day-order_time) == lead_time:

data['orders'].append(q)
inventory += q
order_placed = False
order_time = 0
else:
data['orders'].append(0)

if inventory >= day_demand:

3
 data['units sold'].append(day_demand)
data['units lost'].append(0)
inventory -= day_demand
data['stockouts'].append(0)
else:
data['units sold'].append(inventory)
data['units lost'].append(abs(inventory - day_demand))
inventory = 0
stock_out += 1
data['stockouts'].append(stock_out)

data['current inv level'].append(inventory)

x = pd.DataFrame(data)
return x

data storing

cost function to be optimized

[7]: #"data" being the output of the simulation function "cc_monte_carlo_ray"
def total_cost(data, product):

unit_cost = product.unit_cost

holding_cost = product.holding_cost
order_cost = product.ordering_cost

shortage_cost = unit_cost*0.5

days = max(data['day'])

#service_level = sum(data['units sold']) / sum(data['daily demand'])

no_of_orders = 0

for i in range(days):
if data['orders'][i] != 0:
no_of_orders += 1

Co = no_of_orders * order_cost

#holding cost unit is /unit/day

Ch = sum(data['current inv level']) * holding_cost

4
 Cs = sum(data['units lost'])* shortage_cost

#SLC = service_level*2000

total_cost = Co + Ch + Cs

return total_cost

simulation for no of runs

[8]: def cc_mc_simulation(product, q, r, num_simulations=10):
total_cost_list = []

for sim in range(num_simulations):

data = cc_monte_carlo_ray(product, q, r)

cost = total_cost(data, product)

total_cost_list.append(cost)

return np.mean(total_cost_list)

[ ]:

[9]: product_details.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 400 entries, 0 to 399
Data columns (total 14 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 index 400 non-null int64
1 id 400 non-null int64
2 unit cost 400 non-null int64
3 lead time 400 non-null int64
4 demand probability 400 non-null float64
5 initial inventory 400 non-null int64
6 mean demand 400 non-null int64
7 sd 400 non-null int64
8 demand distribution 400 non-null int64
9 r_opt 400 non-null float64
10 Q_opt 400 non-null float64
11 Min_Cost 400 non-null float64
12 r_pred 400 non-null float64
13 Q_pred 400 non-null float64
dtypes: float64(6), int64(8)
memory usage: 43.9 KB

5
[ ]:

[33]: num_products = len(product_details)

products = []
cost_list2 = []
cost_list = []
avg_cost = []
avg_cost2 = []
r_pred = []
Q_pred = []
r_opt = []
Q_opt = []

for i in range(1, num_products + 1):

product = Product(i)
products.append(product)

for product in products:

r1 = product.pred_r - np.random.randint(15,40)
q1 = product.pred_Q + np.random.randint(7,20)

r2 = product.opt_r
q2 = product.opt_Q

data1 = cc_monte_carlo_ray(product, q1, r1)

z1 = total_cost(data1, product) ;
cost_list.append(z1)

data2 = cc_monte_carlo_ray(product, q2, r2)

z2 = total_cost(data2, product)
cost_list2.append(z2)

y1 = cc_mc_simulation(product, q1, r1)

avg_cost.append(y1)

y2 = cc_mc_simulation(product, q2, r2)

avg_cost2.append(y2)

r_pred.append(r1)
Q_pred.append(q1)
r_opt.append(r2)
Q_opt.append(q2)

[30]: data = {
'r_pred': r_pred,

6
 'Q_pred': Q_pred,
'cost with pred': cost_list,

'r_opt': r_opt,
'Q_opt': Q_opt,

'cost with opt': cost_list2,

'avg cost with pred': avg_cost,

'avg cost with opt': avg_cost2
}

df2 = pd.DataFrame(data)

[12]: df

[12]: r_pred Q_pred cost with pred r_opt Q_opt cost with opt \
0 106.173965 8.208406 1.154961e+06 40.0 126.0 506278.92
1 106.173965 8.208406 1.674384e+06 49.0 47.0 1380473.60
2 106.173965 8.208406 1.951516e+05 29.0 104.0 249603.58
3 106.173965 8.208406 6.750826e+05 29.0 104.0 357067.46
4 104.142456 8.015093 1.820762e+05 24.0 150.0 378088.72
.. … … … … … …
395 97.027977 7.484408 7.132091e+05 29.0 104.0 552407.88
396 97.027977 7.484408 1.226939e+06 36.0 123.0 970416.78
397 106.173965 8.208406 1.659390e+05 23.0 28.0 152052.44
398 106.173965 8.208406 8.320065e+05 29.0 104.0 332603.50
399 121.315689 9.243673 1.913151e+05 37.0 35.0 88401.20

avg cost with pred avg cost with opt

0 1.028101e+06 792354.728
1 1.662146e+06 1366804.784
2 1.885481e+05 248517.352
3 6.977490e+05 521603.722
4 2.143280e+05 386752.380
.. … …
395 7.192943e+05 586835.980
396 1.023174e+06 819615.174
397 1.922973e+05 138466.888
398 8.778066e+05 318922.102
399 1.863168e+05 88192.940

[400 rows x 8 columns]

[37]: df2

7
[37]: r_pred Q_pred cost with pred r_opt Q_opt cost with opt \
0 86.173965 17.208406 8.794216e+05 40.0 126.0 746958.40
1 89.173965 11.208406 1.646663e+06 49.0 47.0 1005690.88
2 89.173965 16.208406 1.445776e+05 29.0 104.0 241830.48
3 86.173965 18.208406 5.665428e+05 29.0 104.0 470804.16
4 83.142456 12.015093 2.187317e+05 24.0 150.0 412546.84
.. … … … … … …
395 78.027977 17.484408 7.309064e+05 29.0 104.0 595549.46
396 76.027977 18.484408 1.014925e+06 36.0 123.0 859572.98
397 84.173965 14.208406 1.479773e+05 23.0 28.0 137613.60
398 84.173965 16.208406 6.014946e+05 29.0 104.0 335810.50
399 100.315689 20.243673 1.810702e+05 37.0 35.0 93828.60

avg cost with pred avg cost with opt

0 1.009346e+06 722408.768
1 1.530879e+06 1387433.368
2 1.531399e+05 243314.106
3 6.715012e+05 531609.892
4 2.151684e+05 388521.608
.. … …
395 7.069004e+05 567727.256
396 1.032817e+06 810695.932
397 1.519768e+05 129910.792
398 6.740883e+05 314828.090
399 1.791593e+05 91326.790

[400 rows x 8 columns]

[13]: df['diff of cost with pred values'] = df['cost with pred']-df['cost with opt']
df['diff of avg cost with pred values'] = df['avg cost with pred']-df['avg cost␣
,→with opt']

[14]: df

[14]: r_pred Q_pred cost with pred r_opt Q_opt cost with opt \
0 106.173965 8.208406 1.154961e+06 40.0 126.0 506278.92
1 106.173965 8.208406 1.674384e+06 49.0 47.0 1380473.60
2 106.173965 8.208406 1.951516e+05 29.0 104.0 249603.58
3 106.173965 8.208406 6.750826e+05 29.0 104.0 357067.46
4 104.142456 8.015093 1.820762e+05 24.0 150.0 378088.72
.. … … … … … …
395 97.027977 7.484408 7.132091e+05 29.0 104.0 552407.88
396 97.027977 7.484408 1.226939e+06 36.0 123.0 970416.78
397 106.173965 8.208406 1.659390e+05 23.0 28.0 152052.44
398 106.173965 8.208406 8.320065e+05 29.0 104.0 332603.50
399 121.315689 9.243673 1.913151e+05 37.0 35.0 88401.20

8
 avg cost with pred avg cost with opt diff of cost with pred values \
0 1.028101e+06 792354.728 648681.916124
1 1.662146e+06 1366804.784 293910.022602
2 1.885481e+05 248517.352 -54451.977832
3 6.977490e+05 521603.722 318015.177733
4 2.143280e+05 386752.380 -196012.496020
.. … … …
395 7.192943e+05 586835.980 160801.182587
396 1.023174e+06 819615.174 256521.929451
397 1.922973e+05 138466.888 13886.514132
398 8.778066e+05 318922.102 499403.035390
399 1.863168e+05 88192.940 102913.921460

diff of avg cost with pred values

0 235745.981107
1 295341.443877
2 -59969.301150
3 176145.298603
4 -172424.422939
.. …
395 132458.364291
396 203558.535172
397 53830.384667
398 558884.517216
399 98123.875597

[400 rows x 10 columns]

[24]: sum(df2['cost with opt'])

[24]: 1093460148.2079995

[25]: sum(df2['cost with pred'])

[25]: 1205091146.210217

[26]: sum(df2['avg cost with opt'])

[26]: 1085136729.1214006

[27]: sum(df2['avg cost with pred'])

[27]: 1204805305.117566

[36]: sum(df2['avg cost with pred'])/sum(df2['avg cost with opt'])

[36]: 1.106233658315364

9
[ ]:

10