If we have changed the source data, then do we have to follow the same step for finding/generating the rules? Or change the method?
I think that you have to generate the rules again. Or may be you can use an incremental method.
It depends on the type of data. For sequential data, you might prefer sequential pattern mining algorithms. You might also transform your data, ignore the timestamps, and apply an itemset mining algorithm.
If your data is not static, there are several cases, including:
-incremental (not too much) updates: then you should appy an incremental algorithm (itemsets, or sequential patterns, depending on your choice).
- streaming data (updates at high rate): then aplpy a streaming solution. It might be based on different model for the "observation window" such as sampling, batches, jumping windows or sliding window.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Graphs