hmma (version 1.1.0)

learnModel: Learna a HMM-A from data

Description

The learnModel function, learns a HMM-A from the supplied data file. The function first creates a random model: the initial and transition distributions are initialized using a Dirichlet https://en.wikipedia.org/wiki/Dirichlet_distribution distribution. Thereafter the model is maximised for the datafile that is supplied.

Usage

learnModel(
  data,
  amountOfStates = 2,
  maxit = 50,
  seed,
  iss = 1e-04,
  debug = FALSE
)

Arguments

data

The datafile. The datafile should be a list containing a dataframe with the data as its $x component and contain the lengths of the observations a the $N component (see details).

amountOfStates

The amount of states.

maxit

The maximum amount of iterations.

seed

Seed (optional).

iss

The Imaginary Sample Size (iss), also called priors, to add data.

debug

Debugmode.

Value

The output of the function is an asymmetric hidden Markov model. This model contains the amount of states, the initial distribution, the transition distribution and the emission distribution (Bayesian networks in the different states).

The model can quicly be visualised with the visualise method. The visualise method does not show the Bayesian networks within the states as this would result in unreadable graphs. Instead, the bnlearn graphviz.plot method can be used (see the examples below).

Details

The learnModel makes use of the mhsmm hmmfit function. An example of the structure of the datafile can be found in hmmaExampleData.

Examples

# NOT RUN {
fit <- learnModel(data = hmmaExampleData, amountOfStates = 3, seed = 1234)
visualise(fit)

# See bn in first state
library(bnlearn)
graphviz.plot(fit$parms.emission[[1]])

# }