Assume I have 4 transactions A,A,A,B. Also assume, A has been grouped with B into a single group. There are similar transactions where say C,C,G,C,C,C and C,C,B,C and C,C,B,C,C are also grouped. Here we see that C has been grouped with B and G but C&B have been grouped twice compared to C&G There are lots of transactions like these.
When a new set of transactions come say C,C,B,A,A,G I need to suggest for each of these transactions, what other transactions it should be paired with. So in this example, the prediction should be as follows.
C -> B (Predicted since it had max probability)
C -> B (Predicted since it had max probability)
B -> C (Predicted)
A -> B (Predicted)
A -> B (Predicted)
G -> C (Predicted)
1) For this simple case what approach/ ML technique is needed to achieve this? 2) If the transactions become more complicated where A,A,A,B,N are grouped, how will I be able to handle this case.
I can go through the transactions and calculate the probability of each transactions in a group, but is there a better approach I should be considering. Are some packages available for these problems?