hmma (version 1.1.0)

visualise: Visualise the HMM-A

Description

The function visualise shows a graphical representation of the found HMM-A. Only the initial distribution and transition distribution are shown. Each state contains a Bayesian network, these would be unreadble when displayed within each state. The bnlearn graphviz.plot method can be used to display the bayesian networks.

Usage

visualise(
  model,
  minProb = 0,
  numDigitsRound = 2,
  relateWidthWithWeight = TRUE,
  widthType = "linear",
  minWidth = 0.05,
  maxWidth = 5
)

Arguments

model

The model.

minProb

The minimum probability of an arc (initial or transition) to be drawn in the visualisation (see details).

numDigitsRound

The amount of decimals that should be presented in the graph.

relateWidthWithWeight

An edge with a high weight will be drawn with a thicker line compared to an edge with a lower weight.

widthType

The thickness function for the lines. Defaults to linear function. When sigmoid is used, a sigmoid function is used to determine the thickness.

minWidth

The minimum thickness for a line. Only used when relateWidthWithWeight is TRUE

maxWidth

The maximum thickness for a line. Only used when relateWidthWithWeight is TRUE

Details

Per default, all transitions (initial and transition) are drawn on the canvas. For larger HMM-As, the visualisation can become cluttered with arcs that have a low probability. By setting the minProb, an arc is only drawn when it is equal to or greater than the specified minProb. The default value for minProb is 0.00 (all arcs are drawn).

To further improve the visualisation, edge line width can be related to the respective weight of the edge. When relateWeightWithThickness is TRUE, an edge with a higher weight will be drawn with a wider line. The minimum width of a line and the maximum width can be supplied via the respective parameters.

The minWidth is associated with a weight of 0.00, the maxWidth is related to the weight 1.00. When widthType is set to 'linear', the width of a line is linearly related to the weight. When widthType is set to 'sigmoid', the width of a line is determined via a sigmoid of the weight.

Examples

# NOT RUN {
# First, we need a model that we want to visualise, we create
# one using the learnModel function.
fit <- learnModel(data = hmmaExampleData, amountOfStates = 3)

# To visualise the states and transitions, we use the visualise method.
# Only lines with a weight of 0.10 are drawn.
visualise(fit, minProb = 0.10)

# When it is not desired to relate the width of a line with its weight,
# this can be disabled:
visualise(fit, minProb = 0.10, relateWidthWithWeight = FALSE)

# Finally, it is possible to use a sigmoid instead of a linear relation:
visualise(fit, minProb = 0.10, widthType = 'sigmoid')

# To visualise the BNs within the states, use the code below
library(bnlearn)
graphviz.plot(fit$parms.emission[[1]])

# }