Question

我有一个数据集（DATE_LOCATION，已售出），其中包含在不同日期销售的产品。这些日期为9个月，每月随机13天或更长时间。我必须以这样的方式分离数据：产品连续销售多少产品1-3天，连续销售4-7天，连续销售8-15天，并连续销售> 16天。那么如何使用pandas和其他包

在python中对此进行编码

      DATE_LOCATION  Sold
      07-08-16 0:00    2
      08-08-16 0:00    7
      12-08-16 0:00    3
      13-08-16 0:00    1
      14-08-16 0:00    2
      15-08-16 0.00    1
      .
      . 
      .
      22-10-16 0:00    1
      23-10-16 0:00    2
      26-10-16 0:00    1
      28-10-16 0:00    1
      29-10-16 0:00    3
      30-10-16 0:00    3
      .
      .
      .(goes for 9 months of data)
      .

我甚至不知道如何在python中编写代码所需的输出是

 Days   Sold
 1-3     20 #(7,8),(22,23),(26),(28,29,30) dates because the range is [1,3]
 4-7      7 #(12,13,14,15) dates because the range is [4,7]
 8-15     0
  >16     0

如果至少有人发布了从哪里开始的链接，那将会很高兴。我试过了

df["DATE_LOCATION"] = pd.to_datetime(df.DATE_LOCATION)
df["DAY"] = df.DATE_LOCATION.dt.day
def flag(x):
    if 1<=x<=3:
        return '1-3'
    elif 4<=x<=7:
        return '4-7'
    elif 8<=x<=15:
        return '8-15'
    else:
        return '>=16'
df["Days"] = df.DAY.apply(flag)
df["Days"].Sold.sum()

这给了我每个月这几天之间销售的产品数量。但是我需要指定范围内的产品总和，其中产品以指定条纹出售。

Answer 1

我通过此代码重现了输入数据

df = pd.DataFrame({'DATE_LOCATION': ['07-08-16 0:00', '08-08-16 0:00', '12-08-16 0:00',\
                                     '13-08-16 0:00', '14-08-16 0:00', '15-08-16 0:00',\
                                     '22-10-16 0:00', '23-10-16 0:00', '26-10-16 0:00',\
                                     '28-10-16 0:00', '29-10-16 0:00', '30-10-16 0:00',],\
                   'Sold': [2, 7, 3, 1, 2, 1, 1, 2, 1, 1, 3, 3]})
df.DATE_LOCATION = pd.to_datetime(df.DATE_LOCATION, dayfirst=True)

现在数据看起来像这样

   DATE_LOCATION  Sold
0     2016-08-07     2
1     2016-08-08     7
2     2016-08-12     3
3     2016-08-13     1
4     2016-08-14     2
5     2016-08-15     1
6     2016-10-22     1
7     2016-10-23     2
8     2016-10-26     1
9     2016-10-28     1
10    2016-10-29     3
11    2016-10-30     3

获取行之间的间隔，计算运行长度（连续天数）并将它们分组，直到运行长度继续扩展，最后得到最大run_length并汇总每个组中已售商品的总和。

df['Day_Interval'] = df.DATE_LOCATION.diff().shift(0).fillna(0)

# calculate run length
day_intervals = (df.Day_Interval.values / np.timedelta64(1, 'D')).astype(int)
run_lengths = []
run_length = 0
groups = []
group = 0

for day_interval in day_intervals:
    if day_interval != 1:
        run_length = 1
        group += 1
        groups.append(group)
    else:
        run_length += 1
        groups.append(group)
    run_lengths.append(run_length)

df['Run_Length'] = run_lengths
df['Group'] = groups

# calculate summary statistic by group
df = df.groupby('Group')['Sold', 'Run_Length'].agg({'Sold': np.sum, 'Run_Length': np.max})
df['1-3'] = 0
df['4-7'] = 0
df['8-15'] = 0
df['>=16'] = 0

df.loc[(df.Run_Length >= 1) & (df.Run_Length <=3), "1-3"] = df.Sold
df.loc[(df.Run_Length >= 4) & (df.Run_Length <=7), "4-7"] = df.Sold
df.loc[(df.Run_Length >= 8) & (df.Run_Length <=15), "8-15"] = df.Sold
df.loc[(df.Run_Length >= 16), ">=16"] = df.Sold
df = df.T.iloc[2:]
df['Sold'] = df.sum(axis=1)
df = df[['Sold']]

输出（df）：

Group   Sold
1-3     20
4-7     7
8-15    0
>=16    0

希望这有帮助。

如何计算某些连续日范围的摘要统计信息

1 个答案: