Fuzzy Hidden Markov Chain with States Depending on Observation for Web Applications
Sharmila V., Sujatha R. and Narasimman S.
This paper provides a novel approach en route to the fuzzy hidden Markov chain to have an additional property namely the observation dependent property. In fuzzy hidden Markov chain the current state depends only on one state that is the immediately preceding state. The newly added property makes the fuzzy hidden Markov chain to depend on two things (1) the immediately preceding state and (2) the immediately preceding observation. The specialism is that though this newly developed fuzzy hidden Markov chain depends on two values the state sequence remains a Markov chain. This paper also solves the three problems evaluation, optimal state sequence and parameter re-estimation for the developed model by giving new algorithms. These algorithms are applied to our institution’s website to know the pattern of the website’s usage.
Keywords: Triangular fuzzy number (TFN), possibility space, conditional possibility, fuzzy Markov chain, fuzzy hidden Markov chain, Viterbi algorithm.