Chapter 7. Sequential Data Models
As seen in the previous chapter, the Naive Bayes model does not make any assumptions regarding the order of events. The likelihood and prior probabilities are computed by aggregating counts. However, some applications such as text or voice recognition, language translation, or decoding, rely on an ordered sequence of states or events [7:1].
Chapter 3, Data Pre-processing, introduced some deterministic solutions to modeling sequences of data, collectively known as time series analysis. An alternative probabilistic solution relies on Markov process and models.
The broad universe of Markov models encompasses computational concepts such as the Markov decision process, discrete Markov, Markov chain Monte Carlo for Bayesian networks, and hidden Markov models.
The first section of this chapter introduces and describes the hidden Markov model along with the dynamic programming techniques used in the evaluation, decoding, and training of models for sequential...
