MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top


I would like to detect repetitive patterns and deviations from these repetitions. I have historical data and can calculate probabilities for the transitions between my many states. I have researched into Markov chains somewhat and feel these could present the correct means of modelling my problem.

However I am confused by this and would appreciate some simple words which I have perhaps missed in the academic literature I have read to date.

Basically if I want to calculate the probability that my current set of events is repetitive so can I simple sum the transitions through my transition matrix and the closer I am to 1 the more certain I am the pattern is repetitive? Do I even need a Markov chains or just the transition matrix?



share|cite|improve this question
This sounds like it might be hard in general. I think you want to express your transition matrix as a weighted sum of permutation matrices with a minimal number of weights providing good approximations. But given this, the coefficients of the permutation matrices would seem to provide a rough measure along these lines. – Steve Huntsman Jul 25 '10 at 13:24
You might also consider the autocorrelation, see e.g. – Steve Huntsman Jul 25 '10 at 14:03
Thanks, I'll look into your suggestions – David Jul 25 '10 at 20:13

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.