CODEWITHSUNDEEP
pythonpandasnumpy
Shawn Nash
Shawn Nash

Reputation: 3

Python Pandas: Merging data frames on multiple conditions

I wish to merge data frames as fetched via sql under multiple condition.

The df1 and df2 are shown below:

df1

Customer ID     Cluster ID  Customer Zone ID
CUS1001.A       CUS1001.X   CUS1000
CUS1001.B       CUS1001.X   CUS1000
CUS1001.C       CUS1001.X   CUS1000
CUS1001.D       CUS1001.X   CUS1000
CUS1001.E       CUS1001.X   CUS1000
CUS2001.A       CUS2001.X   CUS2000

df2:

Complain ID RegistrationNumber   Status
CUS3501.A       99231            open
CUS1001.B       21340            open
CUS1001.X       32100            open

I wish to merge these two data frame with following condition:

if(Complain ID == Customer ID):
    Merge on Customer ID
Elif(Complain ID == Cluster ID):
    Merge on Customer ID
Elif (Complain ID == Customer Zone ID):
    Merge on Customer ID
Else:
    Merge empty row.

Final result should look like this:

Customer ID Cluster ID  Customer Zone ID   Complain ID  Regi ID  Status
CUS1001.A   CUS1001.X       CUS1000         CUS1001.X    32100    open
CUS1001.B   CUS1001.X       CUS1000         CUS1001.B    21340    open
CUS1001.C   CUS1001.X       CUS1000         CUS1001.X    32100    open
  .             .               .               .           .       .
  .             .               .               .           .       .
CUS2001.A   CUS2001.X       CUS2000             0           0       0

Please help!

Upvotes: 0

Views: 2837

Answers (1)

BENY
BENY

Reputation: 323346

Try this ...using pandas: melt, merge and concat

df=pd.melt(df1)
df=df.merge(df2,left_on='value',right_on='Complain ID',how='left')
df['number']=df.groupby('variable').cumcount()
df=df.groupby('number').bfill()
Target=pd.concat([df1,df.iloc[:5,2:6]],axis=1).fillna(0).drop('number',axis=1)

Target
Out[39]: 
  Customer ID Cluster ID Customer Zone ID Complain ID  RegistrationNumber  \
0   CUS1001.A  CUS1001.X          CUS1000   CUS1001.X             32100.0   
1   CUS1001.B  CUS1001.X          CUS1000   CUS1001.B             21340.0   
2   CUS1001.C  CUS1001.X          CUS1000   CUS1001.X             32100.0   
3   CUS1001.D  CUS1001.X          CUS1000   CUS1001.X             32100.0   
4   CUS1001.E  CUS1001.X          CUS1000   CUS1001.X             32100.0   
5   CUS2001.A  CUS2001.X          CUS2000           0                 0.0   
  Status    
0   open         
1   open         
2   open         
3   open         
4   open        
5      0

Update By using numpy's intersect1d, Personally I like this approach most than the previous one . 

df1.MatchId=[np.intersect1d(x,df2.ComplainID.values) for x in df1[['CustomerID','ClusterID']].values]
df1.MatchId=df1.MatchId.apply(pd.Series)
df1
Out[307]:
  CustomerID  ClusterID CustomerZoneID    MatchId
0  CUS1001.A  CUS1001.X        CUS1000  CUS1001.X
1  CUS1001.B  CUS1001.X        CUS1000  CUS1001.B
2  CUS1001.C  CUS1001.X        CUS1000  CUS1001.X
3  CUS1001.D  CUS1001.X        CUS1000  CUS1001.X
4  CUS1001.E  CUS1001.X        CUS1000  CUS1001.X
5  CUS2001.A  CUS2001.X        CUS2000        NaN

df1.merge(df2,left_on='MatchId',right_on='ComplainID',how='left')
Out[311]: 
  CustomerID  ClusterID CustomerZoneID    MatchId ComplainID  \
0  CUS1001.A  CUS1001.X        CUS1000  CUS1001.X  CUS1001.X   
1  CUS1001.B  CUS1001.X        CUS1000  CUS1001.B  CUS1001.B   
2  CUS1001.C  CUS1001.X        CUS1000  CUS1001.X  CUS1001.X   
3  CUS1001.D  CUS1001.X        CUS1000  CUS1001.X  CUS1001.X   
4  CUS1001.E  CUS1001.X        CUS1000  CUS1001.X  CUS1001.X   
5  CUS2001.A  CUS2001.X        CUS2000        NaN        NaN   
   RegistrationNumber Status  
0             32100.0   open  
1             21340.0   open  
2             32100.0   open  
3             32100.0   open  
4             32100.0   open  
5                 NaN    NaN

Upvotes: 1

Related Questions