I want to analyze bank statements to discover useful patterns.
When looking at a person’s bank statements, we find that certain transactions are repeated periodically in a very predictable fashion, like one’s salary or rent, whereas others have a more erratic pattern, such as food expenses, car repairs or vacations. so i want to be able to tell the person what is unnecessary in their transactions.
Which ML/AI techniques can i use and how can i use them?
What are the relevant papers ?
Bayesian networks are a very general tool that can be used for a large number of problems: reasoning (using the Bayesian inference algorithm), learning (using the expectation-maximization algorithm), planning (using decision networks) and perception (using dynamic Bayesian networks). Bayesian networks are used in AdSense to choose what ads to place and on XBox Live to rate and match players.
Probabilistic algorithms can also be used for filtering, prediction, smoothing and finding explanations for streams of data, helping perceptionsystems to analyze processes that occur over time (e.g., hidden Markov models or Kalman filters)
You can try to analyze such data using neural networks or any other ML-techniques. The easiest way to simulate it is using Micosoft Azure ML (try using their free-tier) at https://azure.microsoft.com/en-us/services/machine-learning-studio/
Follow the demo provided after registering an account and you are ready to go.
I have a same problem my undergraduate topic is: automated a bank reconciliation system based on machine learning.
But my concern is bank reconciliation have many steps(7) then is it possible to build a system fully based in machine learning? If yes which algorithm and how to proceed?
How to get a relevant paper about this? I really need help
thank you
Automating any reconciliation is challenging for a number of reasons:
- Scale of matching challenges. Small businesses are easily solved with macros or built in tools that can match on a couple of fields. However, larger enterprises with 100,000's of rows, dirty/missing/inconsistent data breaks rules.
- 1:1 matches are easy, but most reconciliations get into many to one or one to many or many to many scenarios which exponentially add complexity.
- Timing of match periods. Especially true in the investment field which needs matching and posting quickly across time zones and periods.
Sigma IQ (https://sigmaiq.com) is solving this problem through machine learning. There is more info on that here: https://www.sigmaiq.com/finance-blog/automating-bank-reconciliations-through-rpa-machine-learning
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence