Bank Transactions Analysis - Beginner doubt for use cases

Question

Greetings friends!

I am trying to create a model that is able to identify the following patterns:

- For each debited account -> Cluster by transaction values and payment type to check for any patterns.

- For each debited account -> Cluster by Transaction value or payment type and Credited account to check whether for patterns (most used credited account)

I have tried all sort of tools of predictive grouping, but my dataset is too big and I am a beginner on machine learning.

Would love to hear any good ideas to involve other tools as well so I can study. Unfortunatelly there is not much knowledge about this category of tools on the internet.

Thanks

Matthew · Answer

The sample set you provided is much too small for anyone here to help you build a model. But it looks to me like any patterns you find are going to be time series related.. That is unless you have other metadata that an ML algorithm could use

If I were you, I'd look for cycles/trends in the day of week, or week of year.. Perhaps transactions change size seasonally, or maybe the transaction frequency will cluster on certain days/weeks

try grouping your analysis, either by country, or maybe account number (if you have enough transactions)