如何使用 python pandas 结合销售预测、计划交货和当前库存来预测未来库存
How to combine a sales forecast, planned deliveries, and current inventory to project future inventory using python pandas
我有三个数据源：
销售预测 - 这是产品的未来预测销售：
Product
Forecast_Quantity
Forecast_Month
Product A
5
2021-02-28
Product B
6
2021-02-28
Product C
2
2021-02-28
Product A
5
2021-03-31
Product B
6
2021-03-31
Product C
2
2021-03-31
Product A
5
2021-04-30
Product B
6
2021-04-30
Product C
2
2021-04-30
计划交货（采购订单）- 计划交货时间：
Product
Delivery_Quantity
Delivery_Month
Product A
2
2021-02-28
Product B
4
2021-02-28
Product C
5
2021-02-28
Product A
8
2021-03-31
Product B
2
2021-03-31
Product C
4
2021-03-31
Product A
2
2021-04-30
Product B
6
2021-04-30
Product C
3
2021-04-30
当前库存 - 当前库存：
Product
Inventory_Quantity
Inventory_Month
Product A
20
2021-01-31
Product B
16
2021-01-31
Product C
21
2021-01-31
我想创建一个小程序，returns 一个数据框来预测我未来每种产品的库存。它应该需要上个月的结清库存加上交货数量并减去预测的销售额。
预期输出：
Product
Inventory_Quantity
Inventory_Month
Product A
20
2021-01-31
Product B
16
2021-01-31
Product C
21
2021-01-31
Product A
17
2021-02-28
Product B
14
2021-02-28
Product C
24
2021-02-28
这将是 18 个月的数据（但我刚刚为上述 table 使用了 2 个月的输出）。
我已经尝试了几种不同的方法，例如尝试 cusum() 或使用 for 循环，但我的逻辑并不正确。
这是我目前的代码，它将预测和交付组合成一个 'net' 数字：
import pandas as pd
import numpy as np
# Import CSV files as dataframe
fcs = pd.read_csv(r'forecast.csv')
inv = pd.read_csv(r'inventory.csv')
pos = pd.read_csv(r'purchase_orders.csv')
# Inner join to get a net position for each month (PO Qty - Fcs Qty)
net = pd.merge(pos, fcs, left_on=['SKU', 'Delivery_Date'], right_on=['SKU', 'Forecast_Date'])
# Create net
net['net'] = (net['PO_Quantity'] - net['Forecast_Quantity'])
我不确定现在生成与库存结合的新 table 的最佳方法？
使用concat垂直合并三个数据框。然后，fillna 库存列与其他列和 melt 数据框。从那里，您可以乘坐 cumsum:
df = (pd.concat([df1.assign(Forecast_Quantity=df1['Forecast_Quantity'] * -1)
.rename({'Forecast_Month' : 'Inventory_Month',
'Forecast_Quantity' : 'Inventory_Quantity'}, axis=1),
df2.rename({'Delivery_Month' : 'Inventory_Month',
'Delivery_Quantity' : 'Inventory_Quantity'}, axis=1),
df3]).sort_values(['Product', 'Inventory_Month']))
df['Inventory_Quantity'] = df.groupby('Product')['Inventory_Quantity'].cumsum()
df = (df.groupby(['Product', 'Inventory_Month'], as_index=False).last()
.sort_values('Inventory_Month'))
df
Out[1]:
Product Inventory_Month Inventory_Quantity
0 Product A 2021-01-31 20.0
4 Product B 2021-01-31 16.0
8 Product C 2021-01-31 21.0
1 Product A 2021-02-28 17.0
5 Product B 2021-02-28 14.0
9 Product C 2021-02-28 24.0
2 Product A 2021-03-31 20.0
6 Product B 2021-03-31 10.0
10 Product C 2021-03-31 26.0
3 Product A 2021-04-30 17.0
7 Product B 2021-04-30 10.0
11 Product C 2021-04-30 27.0
我有三个数据源：
销售预测 - 这是产品的未来预测销售：
| Product | Forecast_Quantity | Forecast_Month |
|---|---|---|
| Product A | 5 | 2021-02-28 |
| Product B | 6 | 2021-02-28 |
| Product C | 2 | 2021-02-28 |
| Product A | 5 | 2021-03-31 |
| Product B | 6 | 2021-03-31 |
| Product C | 2 | 2021-03-31 |
| Product A | 5 | 2021-04-30 |
| Product B | 6 | 2021-04-30 |
| Product C | 2 | 2021-04-30 |
计划交货（采购订单）- 计划交货时间：
| Product | Delivery_Quantity | Delivery_Month |
|---|---|---|
| Product A | 2 | 2021-02-28 |
| Product B | 4 | 2021-02-28 |
| Product C | 5 | 2021-02-28 |
| Product A | 8 | 2021-03-31 |
| Product B | 2 | 2021-03-31 |
| Product C | 4 | 2021-03-31 |
| Product A | 2 | 2021-04-30 |
| Product B | 6 | 2021-04-30 |
| Product C | 3 | 2021-04-30 |
当前库存 - 当前库存：
| Product | Inventory_Quantity | Inventory_Month |
|---|---|---|
| Product A | 20 | 2021-01-31 |
| Product B | 16 | 2021-01-31 |
| Product C | 21 | 2021-01-31 |
我想创建一个小程序，returns 一个数据框来预测我未来每种产品的库存。它应该需要上个月的结清库存加上交货数量并减去预测的销售额。 预期输出：
| Product | Inventory_Quantity | Inventory_Month |
|---|---|---|
| Product A | 20 | 2021-01-31 |
| Product B | 16 | 2021-01-31 |
| Product C | 21 | 2021-01-31 |
| Product A | 17 | 2021-02-28 |
| Product B | 14 | 2021-02-28 |
| Product C | 24 | 2021-02-28 |
这将是 18 个月的数据（但我刚刚为上述 table 使用了 2 个月的输出）。
我已经尝试了几种不同的方法，例如尝试 cusum() 或使用 for 循环，但我的逻辑并不正确。
这是我目前的代码，它将预测和交付组合成一个 'net' 数字：
import pandas as pd
import numpy as np
# Import CSV files as dataframe
fcs = pd.read_csv(r'forecast.csv')
inv = pd.read_csv(r'inventory.csv')
pos = pd.read_csv(r'purchase_orders.csv')
# Inner join to get a net position for each month (PO Qty - Fcs Qty)
net = pd.merge(pos, fcs, left_on=['SKU', 'Delivery_Date'], right_on=['SKU', 'Forecast_Date'])
# Create net
net['net'] = (net['PO_Quantity'] - net['Forecast_Quantity'])
我不确定现在生成与库存结合的新 table 的最佳方法？
使用concat垂直合并三个数据框。然后，fillna 库存列与其他列和 melt 数据框。从那里，您可以乘坐 cumsum:
df = (pd.concat([df1.assign(Forecast_Quantity=df1['Forecast_Quantity'] * -1)
.rename({'Forecast_Month' : 'Inventory_Month',
'Forecast_Quantity' : 'Inventory_Quantity'}, axis=1),
df2.rename({'Delivery_Month' : 'Inventory_Month',
'Delivery_Quantity' : 'Inventory_Quantity'}, axis=1),
df3]).sort_values(['Product', 'Inventory_Month']))
df['Inventory_Quantity'] = df.groupby('Product')['Inventory_Quantity'].cumsum()
df = (df.groupby(['Product', 'Inventory_Month'], as_index=False).last()
.sort_values('Inventory_Month'))
df
Out[1]:
Product Inventory_Month Inventory_Quantity
0 Product A 2021-01-31 20.0
4 Product B 2021-01-31 16.0
8 Product C 2021-01-31 21.0
1 Product A 2021-02-28 17.0
5 Product B 2021-02-28 14.0
9 Product C 2021-02-28 24.0
2 Product A 2021-03-31 20.0
6 Product B 2021-03-31 10.0
10 Product C 2021-03-31 26.0
3 Product A 2021-04-30 17.0
7 Product B 2021-04-30 10.0
11 Product C 2021-04-30 27.0