如何绘制取决于列值的连续数据

Question

我想要什么

我需要用下面的数据框绘制这样的图表：

0 购买 2021-06-15 0:00:00 60 1 销售 2021-06-17 0:00:00 11 2 销售 2021-06-17 0:00:00 4 3 销售 2021-06-17 0:00:00 2 4 销售 2021-06-18 0:00:00 1 5 销售 2021-06-18 0:00:00 2 6 销售 2021-06-19 0:00:00 1 7 销售 2021-06-21 0:00:00 1 8 销售 2021-06-22 0:00:00 1 9 销售 2021-06-23 0:00:00 1 10 销售 2021-06-25 0:00:00 3 11 销售 2021-06-30 0:00:00 3 12 购买 2021-06-30 0:00:00 50 13 销售 2021-07-01 0:00:00 2 14 销售 2021-07-01 0:00:00 6 15 销售 2021-07-02 0:00:00 1 16 销售 2021-07-02 0:00:00 3 17 销售 2021-07-03 0:00:00 1 18 销售 2021-07-03 0:00:00 2

我试图用cumsum() 进行绘图，但我需要先将带有“销售”的行转换为负值才能做到这一点。

有没有更好的办法？

Answer 1

让我们尝试一下：

# Convert to datetime
df['datetime'] = pd.to_datetime(df['datetime'])
# Negate Sale rows
df.loc[df['action'].eq('Sale'), 'amt'] *= -1
# Calculate the total amount
df['total_amt'] = df['amt'].cumsum()
# Plot datetime vs total_amt
ax = df.plot(x='datetime', y='total_amt', ylabel='Qty', xlabel='Date')
plt.show()

DataFrame 和导入：

import pandas as pd
from matplotlib import pyplot as plt

df = pd.DataFrame({
    'action': ['Buy', 'Sale', 'Sale', 'Sale', 'Sale', 'Sale', 'Sale', 'Sale',
               'Sale', 'Sale', 'Sale', 'Sale', 'Buy', 'Sale', 'Sale', 'Sale',
               'Sale', 'Sale', 'Sale'],
    'datetime': ['2021-06-15 0:00:00', '2021-06-17 0:00:00',
                 '2021-06-17 0:00:00', '2021-06-17 0:00:00',
                 '2021-06-18 0:00:00', '2021-06-18 0:00:00',
                 '2021-06-19 0:00:00', '2021-06-21 0:00:00',
                 '2021-06-22 0:00:00', '2021-06-23 0:00:00',
                 '2021-06-25 0:00:00', '2021-06-30 0:00:00',
                 '2021-06-30 0:00:00', '2021-07-01 0:00:00',
                 '2021-07-01 0:00:00', '2021-07-02 0:00:00',
                 '2021-07-02 0:00:00', '2021-07-03 0:00:00',
                 '2021-07-03 0:00:00'],
    'amt': [60, 11, 4, 2, 1, 2, 1, 1, 1, 1, 3, 3, 50, 2, 6, 1, 3, 1, 2]
})

*列标题被省略，所以我添加了一些。

df.head():

  action            datetime  amt
0    Buy  2021-06-15 0:00:00   60
1   Sale  2021-06-17 0:00:00   11
2   Sale  2021-06-17 0:00:00    4
3   Sale  2021-06-17 0:00:00    2
4   Sale  2021-06-18 0:00:00    1

---

1. 转换日期列to_datetime：

df['datetime'] = pd.to_datetime(df['datetime'])

这将使 xaxis 刻度更易于使用。

1. 否定“action”等于“sales”的行：

df.loc[df['action'].eq('Sale'), 'amt'] *= -1

df.head():

  action   datetime  amt
0    Buy 2021-06-15   60
1   Sale 2021-06-17  -11
2   Sale 2021-06-17   -4
3   Sale 2021-06-17   -2
4   Sale 2021-06-18   -1

1. 用cumsum计算累计总数：

df['total_amt'] = df['amt'].cumsum()

  action   datetime  amt  total_amt
0    Buy 2021-06-15   60         60
1   Sale 2021-06-17  -11         49
2   Sale 2021-06-17   -4         45
3   Sale 2021-06-17   -2         43
4   Sale 2021-06-18   -1         42

4. 然后是DataFrame.plot，x 轴为“datetime”，y 轴为“total_amt”：

ax = df.plot(x='datetime', y='total_amt', ylabel='Qty', xlabel='Date')

---

另一个可选步骤是使用groupby last 仅绘制每天的结束总数：

# Convert to datetime
df['datetime'] = pd.to_datetime(df['datetime'])
# Negate Sale rows
df.loc[df['action'].eq('Sale'), 'amt'] *= -1
# Calculate the total amount
df['total_amt'] = df['amt'].cumsum()
# Get only the Ending Daily Total
daily_df = df.groupby(df['datetime'].dt.date)[['datetime', 'total_amt']].last()
ax = daily_df.plot(x='datetime', y='total_amt', ylabel='Qty', xlabel='Date')
plt.show()

daily_df.head():

             datetime  total_amt
datetime                        
2021-06-15 2021-06-15         60
2021-06-17 2021-06-17         43
2021-06-18 2021-06-18         40
2021-06-19 2021-06-19         39
2021-06-21 2021-06-21         38