firstOrderMarkovModel: A class to train First Order Markov Models

In pMineR: Processes Mining in Medicine

Description

This is an implementation of the First Order Markov Model (FOMM) for Process Mining issues. This class provides a minimal set of methods to handle with the FOMM model:

• firstOrderMarkovModel( ) is the costructor

• loadDataset( ) loads data taken from a dataLoader::getData() method, into a FOMM object

• trainModel( ) train a model using the previously loaded dataset

• replay( ) re-play a given event log on the internal FOMM model

• play( ) play the internal FOMM model a desired number of times, in order to simulate new event-logs. This methods can also, if desired, simulate event-logs which does not complies with the internal FOMM model.

• plot( ) plot the internal model

• distanceFrom( ) calculate the scalar distance to another passed FOMM model, passed as argument. The default metric returns a scalar value

• getModel( ) return the trained internal FOMM model

• getLogObj( ) return an XML containing the execution-log of a firstOrderMarkovModel::play() computation

• getInstanceClass( ) return the instance class Name and description (version, etc.)

• plot.delta.graph( ) plot a graph, in the desired modality, representing the difference between the internal FOMM and a passed one.

• get.transition.Prob( ) calculate the probability to go in a given number of transitions, from a state to another

• get.time.transition.Prob( ) calculate the probability to go in a given time, from a state to another

• build.PWF( ) build automatically a PWF XML definition script.

• findReacheableNodes( ) and return the array containing the reacheable states, starting from the passed one.

In order to better undestand the use of such methods, please visit: www.pminer.info

The consturctor admit the following parameters: parameters.list a list containing possible parameters to tune the model.

Usage

firstOrderMarkovModel(parameters.list = list())

Arguments

parameters.list

a list containing the parameters. The possible ones are: 'considerAutoLoop' and 'threshold'. 'considerAutoLoop' is a boolean which indicates if the autoloops have to be admitted, while 'threshold' is the minimum value that a probability should have to do not be set to zero, in the transition matrix.

Examples

## Not run: 

# create a Loader
obj.L<-dataLoader();   

# Load a .csv 
obj.L$load.csv(nomeFile = "../otherFiles/mammella.csv",
IDName = "CODICE_SANITARIO_ADT",
EVENTName = "DESC_REPARTO_RICOVERO",
dateColumnName = "DATA_RICOVERO")

# get the loaded data
dati <- obj.L$getData()

# build a Second Order Marvov Model with a threshold of 0.2
FOMM <- firstOrderMarkovModel( 
parameters.list = list("threshold"=0.001))

# load the data
FOMM$loadDataset(dataList = dati)

# train a model
FOMM$trainModel()

# generate 10 new processes (nb: if the 
# threshold is too low, it can fail...)
aaa <- FOMM$play(numberOfPlays = 10)

# get the transition matrix
TranMatrix <- FOMM$getModel(kindOfOutput = "MMatrix.perc")

# plot the model
FOMM$plot()

# generate other 20 fake-processes
ooo <- FOMM$play(numberOfPlays = 20)


## End(Not run)

pMineR documentation built on May 2, 2019, 9:34 a.m.