我有一个庞大的dataframe
。结构数据如下:
df
ID Annotation X Y
A Boarding 767513.9918 9425956.2571
A Alighting 767154.1396 9427584.0004
B Boarding 767450.5277 9432627.9543
B Alighting 767495.0101 9426797.1772
C Boarding 767648.9507 9426442.5497
C Alighting 767037.0309 9428878.9032
........
X和Y数据使用UTM坐标。我想计算每个ID登机与下车之间的距离。我的问题很安静,但与此问题(Distance matrix in Python Pandas)不同。我的预期结果如下:
result
ID Anotation_1 X_1 Y_1 Anotation_2 X_2 Y_2 Dist
A Boarding 767513.99 9425956.26 Alighting 767154.14 9427584.00 1667.05
B Boarding 767450.53 9432627.95 Alighting 767495.01 9426797.18 5830.95
C Boarding 767648.95 9426442.55 Alighting 767037.03 9428878.90 2512.02
.......
谢谢您的帮助。
答案 0 :(得分:2)
我将旋转数据框:
result = df.pivot('ID', 'Annotation', ['X', 'Y'])
获得
X Y
Annotation Alighting Boarding Alighting Boarding
ID
A 767154.1396 767513.9918 9.427584e+06 9.425956e+06
B 767495.0101 767450.5277 9.426797e+06 9.432628e+06
C 767037.0309 767648.9507 9.428879e+06 9.426443e+06
然后,我将重命名列并重新索引:
ix = result.columns.to_frame()
result.columns = ix['Annotation'] + '_' + ix.iloc[:,0]
result = result.reindex(columns=['Alighting_X', 'Alighting_Y', 'Boarding_X', 'Boarding_Y'])
获得:
Alighting_X Alighting_Y Boarding_X Boarding_Y
ID
A 767154.1396 9.427584e+06 767513.9918 9.425956e+06
B 767495.0101 9.426797e+06 767450.5277 9.432628e+06
C 767037.0309 9.428879e+06 767648.9507 9.426443e+06
现在很容易计算距离:
result['Dist'] = np.sqrt((result.Alighting_X - result.Boarding_X)**2 + (result.Alighting_Y - result.Boarding_Y)**2)
最终得到:
Alighting_X Boarding_X Alighting_Y Boarding_Y Dist
ID
A 767154.1396 767513.9918 9.427584e+06 9.425956e+06 1667.045847
B 767495.0101 767450.5277 9.426797e+06 9.432628e+06 5830.946773
C 767037.0309 767648.9507 9.428879e+06 9.426443e+06 2512.023929
答案 1 :(得分:2)
我正在使用unstack()
:
m=(df.assign(k=(df.groupby('ID').cumcount()+1).astype(str)).
set_index(['ID','k']).unstack().sort_values(by='k',axis=1))
m.columns=m.columns.map('_'.join)
m=m.assign(Dist=np.sqrt((m.X_1 - m.X_2)**2 + (m.Y_1 - m.Y_2)**2))
print(m)
答案 2 :(得分:1)
一种解决方法,假设输入是正确且正确的,将使用groupby
:
df = df.groupby('ID').apply(lambda x: pd.Series(x.values[0:2,2:4].flatten())) # (*)
df.columns=['X_1','Y_1','X_2','Y_2']
#df.reset_index() # Uncomment if you want 'ID' as a column and not an Index
至于您希望得到的结果中的其他列:Anotation_1
和Anotation_2
始终是恒定的,因此我不必费心将它们包括在内。 Dist
列-好吧,有了新的列,您现在就可以计算出它,或者您可以更改上面的代码来计算距离,同时遍历上面的步骤(*)
中的数字,从而更改代码类似:(此处使用了虚拟距离计算,请用您的虚拟距离代替!)
def my_func(pdf):
return pd.Series([pdf.values[0,2], pdf.values[0,3], pdf.values[1,2], pdf.values[1,3],
np.sqrt((pdf.values[0,2]-pdf.values[1,2])**2+(pdf.values[0,3]-pdf.values[1,3])**2) # <= your distance calculation goes here...
])
df = df.groupby('ID').apply(my_func)
df.columns=['X_1','Y_1','X_2','Y_2','Dist']
#df.reset_index() # Uncomment if you want 'ID' as a column and not an Index
更新:如果您坚持要包含这些常量列,则可以稍后像这样简单地添加它们:(但是为什么会这样呢?尤其是当DataFrame
... 很大时)
df['Annotation_1'] = 'Boarding'
df['Annotation_2'] = 'Alighting'
# And if you further insist on a specific ordering of the columns, you can go with:
df = df[['Annotation_1', 'X_1', 'Y_1', 'Annotation_2', 'X_2', 'Y_2', 'Dist']]