Dear all,
I have been doing tests using SVM, random forests and PPR. The data is from a
data stream (that is, the data for training and for test is always increasing /
changing). With SVM and random forests everything is ok, but with ppr there are
situations where it crashes. For the examples I have used I noticed that if one
of the variables has just one value (it can happen), it crashes for some
parameters set. If there are 2 variables with the same behavior (for example: v1
<- c('A', 'A',
'B','C','B','A') and v2 <- c(1,1,2,3,2,1),
for some parameters sets it crashes, also.
One example:
1. Copy the following data to a new text file named 'errorexample.txt'.
"InicioViagem","Duracao","DiaSemana","DiaAno"
"1705",47364,4376,"quarta-feira ",56
"1695",47900,5270,"quarta-feira ",56
"1692",49828,5182,"quarta-feira ",56
"1690",50779,5489,"quarta-feira ",56
"1703",52297,5712,"quarta-feira ",56
"1717",47157,4867,"quinta-feira ",57
"1735",47985,4669,"quinta-feira ",57
"1708",48425,4760,"quinta-feira ",57
"1709",49806,4703,"quinta-feira ",57
"1715",51518,5931,"quinta-feira ",57
"1722",52853,4966,"quinta-feira ",57
"1732",55174,5560,"quinta-feira ",57
"1706",55763,6023,"quinta-feira ",57
"1718",57362,4969,"quinta-feira ",57
"1736",58201,5718,"quinta-feira ",57
"1710",60215,5130,"quinta-feira ",57
"1723",63349,4754,"quinta-feira ",57
"1733",65485,5080,"quinta-feira ",57
"1750",50453,4984,"sexta-feira ",58
"1748",50985,5409,"sexta-feira ",58
"1741",51490,4952,"sexta-feira ",58
"1761",52761,5253,"sexta-feira ",58
"1737",56321,5326,"sexta-feira ",58
"1756",57125,5113,"sexta-feira ",58
"1738",58168,5944,"sexta-feira ",58
"1746",59392,5256,"sexta-feira ",58
"1751",60682,5962,"sexta-feira ",58
"1749",61187,5705,"sexta-feira ",58
"1762",62973,4829,"sexta-feira ",58
2. Running the next code:
data <- read.table('errorexample.txt',
header=TRUE,sep=",",dec=".")
learner <- 'ppr'
learner.pars <- list(Duracao ~ ., nterms=1, data=data,
kernel='supsmu', na.action=na.omit, max.terms=5, optlevel=1,
sm.method='supsmu', bass=0, span=0)
model <- do.call(learner, learner.pars)
It works fine.
3. But running:
data <- read.table('errorexample.txt',
header=TRUE,sep=",",dec=".")
learner <- 'ppr'
learner.pars <- list(Duracao ~ ., nterms=1, data=data,
kernel='supsmu', na.action=na.omit, max.terms=5, optlevel=2,
sm.method='supsmu', bass=0, span=0)
model <- do.call(learner, learner.pars)
It crashes.
The only difference is to change optlevel from 1 to 2. This is just one example.
It crashes with other examples. Those problems do not allow the use of ppr in a
dynamic environment.
I am very interested in the solution of this problem, If I can help anyway,
please tell me how.
Thank you
Joao Moreira
[[alternative HTML version deleted]]
