I'm trying to find recurring patterns in a stream of bits in order to construct a Gilbert-Eliot model for packet losses and retransmissions in a wireless network. The Model requires interesting patterns to calculate the transition probabilities between states.
Problem: Let's say I have a stream of bits (01000111011101101101...) and I would like to learn/find the X most recurring patterns (for 1).
- pattern --> no
- 11 --> 6
- 101 --> 4
- 111 --> 2
I have a couple of MB of binary data so it is almost impossible to check for interesting recurring patterns by hand.
I'm new to machine learning and data-mining and I was wondering if someone has an idea of what kind of algorithm to use for this. I've looked at association rule learning but this seems a bit overkill for the results I would like to achieve.
Any tips or tricks are welcome.
Thank you very much for the help!