我有一个如下所示的df
B_ID No_Show Session slot_num Cumulative_no_show
1 0.4 S1 1 0.4
2 0.3 S1 2 0.7
3 0.8 S1 3 1.5
4 0.3 S1 4 1.8
5 0.6 S1 5 2.4
6 0.8 S1 6 3.2
7 0.9 S1 7 4.1
8 0.4 S1 8 4.5
9 0.6 S1 9 5.1
12 0.9 S2 1 0.9
13 0.5 S2 2 1.4
14 0.3 S2 3 1.7
15 0.7 S2 4 2.4
20 0.7 S2 5 3.1
16 0.6 S2 6 3.7
17 0.8 S2 7 4.5
19 0.3 S2 8 4.8
在df上方创建的代码如下所示。
import pandas as pd
import numpy as np
df = pd.DataFrame({'B_ID': [1,2,3,4,5,6,7,8,9,12,13,14,15,20,16,17,19],
'No_Show': [0.4,0.3,0.8,0.3,0.6,0.8,0.9,0.4,0.6,0.9,0.5,0.3,0.7,0.7,0.6,0.8,0.3],
'Session': ['s1','s1','s1','s1','s1','s1','s1','s1','s1','s2','s2','s2','s2','s2','s2','s2','s2'],
'slot_num': [1,2,3,4,5,6,7,8,9,1,2,3,4,5,6,7,8],
})
df['Cumulative_no_show'] = df.groupby(['Session'])['No_Show'].cumsum()
和一个名为walkin_no_show = [0.3、0.4、0.3、0.4、0.3、0.4等等，长度为1000的列表]
使用upcumulative> 0.8从上方创建一个新行，紧随其后
df[No_Show] = walkin_no_show[i]
及其会话和slot_num应该与前一个相同，并通过从前一个减去（1-walkin_no_show [i]）创建一个名为u_cumulative的新列。
预期输出：
B_ID No_Show Session slot_num Cumulative_no_show u_cumulative
1 0.4 S1 1 0.4 0.4
2 0.3 S1 2 0.7 0.7
3 0.8 S1 3 1.5 1.5
walkin1 0.3 S1 3 1.5 0.8
4 0.3 S1 4 1.8 1.1
walkin2 0.4 S1 4 1.8 0.5
5 0.6 S1 5 2.4 1.1
walkin3 0.3 S1 5 2.4 0.4
6 0.8 S1 6 3.2 1.2
walkin4 0.4 S1 6 3.2 0.6
7 0.9 S1 7 4.1 1.5
walkin5 0.3 S1 7 4.1 0.8
8 0.4 S1 8 4.5 1.2
walkin6 0.4 S1 8 4.5 0.6
9 0.6 S1 9 5.1 1.2
12 0.9 S2 1 0.9 0.9
walkin1 0.3 S2 1 0.9 0.2
13 0.5 S2 2 1.4 0.7
14 0.3 S2 3 1.7 1.0
walkin2 0.4 S2 3 1.7 0.4
15 0.7 S2 4 2.4 1.1
walkin3 0.3 S2 4 2.4 0.4
20 0.7 S2 5 3.1 1.1
walkin4 0.4 S2 5 3.1 0.5
16 0.6 S2 6 3.7 1.1
walkin5 0.3 S2 6 3.7 0.4
17 0.8 S2 7 4.5 1.2
walkin6 0.4 S2 7 4.5 0.6
19 0.3 S2 8 4.8 0.9
我尝试了以下代码小的修改。正如@ Ben.T在下面回答的那样，提到了我的问题。
create new rows based the values of one of the column in pandas or numpy
感谢@ Ben.T。全部归功于您。.
def create_u_columns (ser):
l_index = []
arr_ns = ser.to_numpy()
# array for latter insert
arr_idx = np.zeros(len(ser), dtype=int)
walkin_id = 1
for i in range(len(arr_ns)-1):
if arr_ns[i]>0.8:
# remove 1 to u_no_show
arr_ns[i+1:] -= (1-walkin_no_show[arr_idx])
# increment later idx to add
arr_idx[i] = walkin_id
walkin_id +=1
#return a dataframe with both columns
return pd.DataFrame({'u_cumulative': arr_ns, 'mask_idx':arr_idx}, index=ser.index)
df[['u_cumulative', 'mask_idx']]= df.groupby(['Session']['Cumulative_no_show'].apply(create_u_columns)
# select the rows
df_toAdd = df.loc[df['mask_idx'].astype(bool), :].copy()
# replace the values as wanted
df_toAdd['No_Show'] = walkin_no_show[mask_idx]
df_toAdd['B_ID'] = 'walkin'+df_toAdd['mask_idx'].astype(str)
df_toAdd['u_cumulative'] -= 1
# add 0.5 to index for later sort
df_toAdd.index += 0.5
new_df_0.8 = pd.concat([df,df_toAdd]).sort_index()\
.reset_index(drop=True).drop('mask_idx', axis=1)
我也想遍历一个列表。我们可以在其中更改（arr_ns [i]> 0.8）[0.8，0.9，1.0]并创建3 df，例如new_df_0.8，new_df_0.9和new_df_1.0
答案 0 :(得分：1)
您唯一需要考虑的技巧是增加索引值的方式。 这是一个解决方案：
walkin_no_show = [0.3，0.4，0.3，0.4，0.3]
TypeError: The added layer must be an instance of class Layer. Found: <keras.layers.core.Dense object at 0x7f736a163e10>
输出：
df = pd.DataFrame({'B_ID': [1,2,3,4,5],
'No_Show': [0.1,0.1,0.3,0.5,0.6],
'Session': ['s1','s1','s1','s2','s2'],
'slot_num': [1,2,3,1,2],
'Cumulative_no_show': [1.5, 0.4, 1.6, 0.3, 1.9]
})
df = df[['B_ID', 'No_Show', 'Session', 'slot_num', 'Cumulative_no_show']]
df['u_cumulative'] = df['Cumulative_no_show']
print(df.head())
然后：
B_ID No_Show Session slot_num Cumulative_no_show u_cumulative
0 1 0.1 s1 1 1.5 1.5
1 2 0.1 s1 2 0.4 0.4
2 3 0.3 s1 3 1.6 1.6
3 4 0.5 s2 1 0.3 0.3
4 5 0.6 s2 2 1.9 1.9
输出：
def Insert_row(row_number, df, row_value):
# Starting value of upper half
start_upper = 0
# End value of upper half
end_upper = row_number
# Start value of lower half
start_lower = row_number
# End value of lower half
end_lower = df.shape[0]
# Create a list of upper_half index
upper_half = [*range(start_upper, end_upper, 1)]
# Create a list of lower_half index
lower_half = [*range(start_lower, end_lower, 1)]
# Increment the value of lower half by 1
lower_half = [x.__add__(1) for x in lower_half]
# Combine the two lists
index_ = upper_half + lower_half
# Update the index of the dataframe
df.index = index_
# Insert a row at the end
df.loc[row_number] = row_value
# Sort the index labels
df = df.sort_index()
# return the dataframe
return df
walkin_count = 1
skip = False
last_Session = ''
i = 0
while True:
row = df.loc[i]
if row['Session'] != last_Session:
walkin_count = 1
last_Session = row['Session']
values_to_append = ['walkin{}'.format(walkin_count), walkin_no_show[i],
row['Session'], row['slot_num'], row['Cumulative_no_show'], (1 - walkin_no_show[i])]
if row['Cumulative_no_show'] > 0.8:
df = Insert_row(i+1, df, values_to_append)
walkin_no_show.insert(i+1, 0)
walkin_count += 1
i += 1
i += 1
if i == df.shape[0]:
break
print(df)
我希望这会有所帮助。
导入的已使用函数答案 1 :(得分：1)
IIUC，您可以通过以下方式实现：
def create_u_columns (ser, threshold_ns = 0.8):
arr_ns = ser.to_numpy()
# array for latter insert
arr_idx = np.zeros(len(ser), dtype=int)
walkin_id = 0 #start at 0 not 1 for list indexing
for i in range(len(arr_ns)-1):
if arr_ns[i]>threshold_ns:
# remove 1 to u_no_show
arr_ns[i+1:] -= (1-walkin_no_show[walkin_id]) #this is slightly different
# increment later idx to add
arr_idx[i] = walkin_id+1
walkin_id +=1
#return a dataframe with both columns
return pd.DataFrame({'u_cumulative': arr_ns, 'mask_idx':arr_idx}, index=ser.index)
#create empty dict for storing the dataframes
d_dfs = {}
#iterate over the value for the threshold
for th_ns in [0.8, 0.9, 1.0]:
#create a copy and do the same kind of operation
df_ = df.copy()
df_[['u_cumulative', 'mask_idx']]= \
df_.groupby(['Session'])['Cumulative_no_show']\
.apply(lambda x: create_u_columns(x, threshold_ns=th_ns))
# select the rows
df_toAdd = df_.loc[df_['mask_idx'].astype(bool), :].copy()
# replace the values as wanted
df_toAdd['No_Show'] = np.array(walkin_no_show)[df_toAdd.groupby('Session').cumcount()]
df_toAdd['B_ID'] = 'walkin'+df_toAdd['mask_idx'].astype(str)
df_toAdd['u_cumulative'] -= (1 - df_toAdd['No_Show'])
# add 0.5 to index for later sort
df_toAdd.index += 0.5
d_dfs[th_ns] = pd.concat([df_,df_toAdd]).sort_index()\
.reset_index(drop=True).drop('mask_idx', axis=1)
然后，如果要访问数据框，则可以执行以下操作：
for th, df_ in d_dfs.items():
print (th)
print (df_.head(4))