pMineR (version 0.31)

firstOrderMarkovModel: A class to train First Order Markov Models

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)