Intelligent Methods for Estimating Hidden Markov Model Parameters

Abstract

Hidden Markov models (HMMs) are stochastic models that were initially applied as statistical models for speech and handwriting recognition because of their great ability to adapt to the problem and their skill in dealing with sequential signals. With the development of techniques, tools, and methods for estimating the parameters of the hidden Markov model, attention may turn to smart methods due to their importance and wide use among researchers. Two algorithms were taken, namely the Ant Colony Optimization (ACO) algorithm and the Bees Colony Optimization (BCO) algorithm. They were applied to an important field at present, which is cyber security, where we addressed one of the threats that pose a danger, which is the Advanced Persistent Threat (APT). The results showed the flexibility in dealing with this type of algorithm for cyber security problems by clarifying the nature of the transitions between the two model states and the emissions that come from each hidden state. It is worth noting that a comparison of the results was made using the two comparison standards, Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC), and we found that the best method was the results of the bee colony algorithm.