Pattern Clustering using Python

user_17778275 0
Viewed 68

I wanted to recognise all the patterns data can be present in "Subset" Column of following data frame.

DataFrame-

Name    Subset    
A-1001  0
A-1001  1
A-1001  2
B-1005  3
B-1005  4
D-1015  0
D-1015  1
D-1015  2
L-650   0 
L-650   5
L-650   6
V-895   3
V-895   4

Here in this data frame, Patterns for Subsets of A-1001, D-1015 Are matching, while B-1005, V-895 have same pattern. Pattern for L-650 is different.

i.e Output Example

Pattern#    Name
1           A-1001, D-1015
2           B-1005, V-895
3           L-650

How can I recognize these patterns using Python?

P.S There may be many unknown patterns.

hierarchical-clustering pandas pattern-matching python
1 Answers

IIUC, you want to perform a double groupby, once to aggregate the Subsets per name and form "patterns", and a second time to aggregate the patterns to group the Names:

import numpy as np

(df.groupby('Name', as_index=False)
   ['Subset'].agg(frozenset)      # use a tuple instead of frozenset if order matters
   .groupby('Subset', as_index=False)
   .agg(list)                     # use ', '.join instead of list to have a string
   .assign(Pattern=lambda d: np.arange(len(d))+1)
)

output:

      Subset              Name  Pattern
0  (0, 1, 2)  [A-1001, D-1015]        1
1     (3, 4)   [B-1005, V-895]        2
2  (0, 5, 6)           [L-650]        3
user_163434640
Related