nb_concept_drift_detector_from_stream: Naive Bayes concept drift detector from an AMIDST data stream

In ramidst: An Interface to the AMIDST Toolbox for Data Stream Processing

Description

Naive Bayes concept drift detector from an AMIDST data stream

Usage

nb_concept_drift_detector_from_stream(input_stream, class_index = -1L,
  window_size, transition_variance = 0.1, hidden_vars = 1L)

Arguments

input_stream	an AMIDST input stream
class_index	the index of the class variable in the list of variables
window_size	the number of items in the stream to be analysed simultaneously
transition_variance	the variance of the transition distribution
hidden_vars	the number of global hidden variables to include in the model

Value

the value of the hidden variables for each window

Note

The function builds a dynamic naive Bayes model with a Gaussian hidden variable which is aimed at capturing an underlying unobserved process.

References

H. Borchani, A.M. Martinez, A.R. Masegosa, H. Langseth, T.D. Nielsen, A. Salmeron, A. Fernandez, A.L. Madsen, R.Saez (2015) Modeling concept drift: A probabilistic graphical model based approach. IDA'2015. Lecture Notes in Computer Science 9385, 72-83.

Examples

## Not run: 
data <- amidst_data_stream(system.file("extdata","sea.arff",
package="ramidst"))
results <- nb_concept_drift_detector_from_stream(data,class_index = -1L,
window_size=1000L,transition_variance=0.1,hidden_vars=1L)
re <- 0
for (k in 1:length(results)) re[k] <- results[[k]]
ymin = min(re)-0.05
ymax = max(re)+0.05
plot(re,type="l",ylim=c(ymin,ymax),ylab="Hidden variable",
xlab="Instance number (x 1000)")
abline(v=15,col="red")
abline(v=30,col="red")
abline(v=45,col="red")

## End(Not run)

ramidst documentation built on May 30, 2017, 5:27 a.m.