How to find behaviour patterns in banking transactional level data and how to extract rules of behaviour



Hi All,

Currently we are working on transactional level of data of customers,which has good or bad flag mapped to it.
How to find behaviour patterns in data and how to extract rules of behaviour?
For this kindly suggest suitable machine learning algorithm/technique.


Hi @sanjana_jha,

If your purpose if the extract best rules from a model, then “decision tree” algorithm would be your best bet, because it is the most “white box” ML algorithm, i.e. you can extract rules from the model after training



Use Decision Tree for modeling purpose…

One Link for Data Mining and pattern Discovery , i hope it will help you.



Thanks for the response. so data is quite imbalanced. Only 2% of data is bad flag. I used smote technique to balance the class and ran random forest. After that fetched the rules from that. But as smote techniques changes the values of features,rules that are being captured have values which are absolutely different from the real raw data.Hence,not able to read the rules and quantify it.
Can someone please help.
Thanks in advance.


Hi @sanjana_jha

You might be interested to check this machine learning project on imbalanced data:

Let me know if this doesn’t answer your question.