synthetic.tseries {TSclust}

Synthetic Time Series for Clustering Performace Comparisons.

Description

This dataset features three repetitions of several models of time series.

Usage

data(synthetic.tseries)

Details

The dataset is a mts object, formed by several repetitions of each of the following models.

M1	AR	X_t = 0.6 X_{t-1} + \varepsilon_{t}
M2	Bilinear	X_t = \left( 0.3 -0.2 \varepsilon_{t-1} \right) X_{t-1} + 1.0 +\varepsilon_{t}
M3	EXPAR	X_t =\left( 0.9 \exp \left( - X_{t-1}^2 \right) -0.6 \right) X_{t-1} + 1.0 + \varepsilon_{t}
M4	SETAR	X_t =\left( 0.3 X_{t-1} +1.0 \right) I \left( X_{t-1} \geq 0.2 \right) -
		\left( 0.3 X_{t-1} -1.0 \right) I \left( X_{t-1} < 0.2 \right) + \varepsilon_{t}
M5	NLAR	X_t = 0.7 \left| X_{t-1} \right| \left( 2 + \left| X_{t-1} \right| \right)^{-1} + \varepsilon_{t}
M6	STAR	X_t = 0.8 X_{t-1} -0.8 X_{t-1} \left( 1 + \exp \left( -10 X_{t-1} \right) \right)^{-1} + \varepsilon_{t}

Three simulations of each model are included. This dataset can be used for comparing the performance of different dissimilarity measures between time series or clustering algorithms.

References

Montero, P and Vilar, J.A. (2014) TSclust: An R Package for Time Series Clustering. Journal of Statistical Software, 62(1), 1-43. http://www.jstatsoft.org/v62/i01/.

Examples

data(synthetic.tseries)
#Create the true solution, for this dataset, there are three series of each model
true_cluster <- c(1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 4, 5, 5, 5, 6, 6, 6)
#test a dissimilarity metric and a cluster algorithm
intperdist <- diss( synthetic.tseries, "INT.PER") #create the distance matrix
#use hierarchical clustering and divide the tree in 6 clusters
intperclust <- cutree( hclust(intperdist), 6 ) 
#use a cluster simmilarity index to rate the solution
cluster.evaluation( true_cluster, intperclust)

#test another dissimilarity metric and a cluster algorithm
acfdist <- diss( synthetic.tseries, "ACF", p=0.05) 
acfcluster <- pam( acfdist, 6 )$clustering #use pam clustering to form 6 clusters
cluster.evaluation( true_cluster, acfcluster)

#test another dissimilarity metric and a cluster algorithm
chernoffdist <- diss( synthetic.tseries, "SPEC.LLR")
chernoffclust <- pam( chernoffdist, 6 )$clustering 
cluster.evaluation( true_cluster, chernoffclust)