I am having some events which are time dependent. For example, a malfunctioning of a component caused substantial change in some parameters and that may lead to another event after a certain amount of time interval, say 1min. I have to classify the upcoming event from a group of events prior to its occurrence using fuzzy rule base system based on the change of the parameters after the malfunctioning of the component and before occurrence of the event.
I think fuzzy rule based system in these case may not be much relevant. b'coz what you are talking about is sequence of events, which are dependent. Fuzzy Rule may not be able to model these dependency automatically from data (you have manually code it in your rule base, I guess). according to me the better idea would be to use some sort of probabilistic models, like Hidden Markov Models, CRF or Bayesian Network, which will help you answer the queries like "What is the probability of an event D occurring, given that some events A, B, C (or may be more) has already occurred?".
there are many successful application of HMM and CRF in sequence prediction and modelling like NPL, Speech, Text processing, Bio-informatics etc.
Dear Mukesh, the information was really informative. The probabilistic models which you have mentioned about is totally new for me. So would you please guide me as to which documents or source of information would be most favorable to know the basics of these models.
- Badges
- Science topic