Regularization for variable selection in model-based clustering and discriminant analysis
SelvarMix is a package where a regularization approach of variable selection is considered in model-based clustering and discriminant analysis frameworks. First, this procedure consists of ranking the variables with a lasso-like procedure. Second, the method of Maugis et al (2009, 2011) is adapted to define the role of variables in the two frameworks. SelvarMix provides a faster variable selection algorithm than the backward stepwise or forward stepwise algorithms of Maugis et al (2009), allowing us to study high-dimensional datasets.
|License:||GPL-3 + file LICENSE|
The general purpose of the package is to perform variable selection in model-based clustering and discriminant analysis. It focus on model-based clustering, where the clusters are assumed to arise from Gaussian distributions. The most achieved model in model-based clustering has been proposed by Maugis et al (2009). This so-called SRUW modeling considers three roles of variables: one variable my belong to the relevant clustering set S, the redundant variable set U or the independent variable set W. Moreover, the redundant variables may be explained by a subset R of the relevant variables S. In order to avoid the slow of this algorithm when data with numerous variables are studied, the SelvarMix procedure is proposed. It proceeds in two steps: First, the variables are ranked using a lasso-like procedure analogous to the one of Zhou et al (2009); second, the SRUW procedure is run on this ranked set of variables.
Author: Mohammed Sedki, Gilles Celeux and Cathy Maugis-Rabusseau
Maugis, C., Celeux, G., and Martin-Magniette, M. L., 2009. "Variable selection in model-based clustering: A general variable role modeling". Computational Statistics and Data Analysis, vol. 53/11, pp. 3872-3882.
Maugis, C., Celeux, G., and Martin-Magniette, M. L., 2011. "Variable selection in model-based discriminant analysis". Journal of Multivariate Analysis, vol. 102, pp. 1374-1387.
Zhou, H., Pan, W., and Shen, X., 2009. "Penalized model-based clustering with unconstrained covariance matrices". Electronic Journal of Statistics, vol. 3, pp.1473-1496.
Sedki, M., Celeux, G., Maugis-Rabusseau, C., 2014. "SelvarMix: A R package for variable selection in model-based clustering and discriminant analysis with a regularization approach". Inria Research Report available at http://hal.inria.fr/hal-01053784
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
## Not run: ## Simulated data example as shown in Maugis et al. (2009) (correlated scenario 2) ## n = 2000 observations, p = 14 variables require(Rmixmod) require(glasso) data(scenarioCor) data.cor <- scenarioCor[,1:14] labels.cor <-scenarioCor[,15] lambda <- seq(20, 100, by = 10) rho <- seq(1, 2, length=2) hybrid.size <- 3 models <- mixmodGaussianModel(family = "spherical", equal.proportions = TRUE) regModel <- c("LI","LB","LC") indepModel <- c("LI","LB") ## variable selection in model-based clustering nbCluster <- c(3,4) criterion <- "BIC" simulate.cl <- SelvarClustLasso(data.cor, nbCluster, lambda, rho, hybrid.size, criterion, models, regModel, indepModel) ## variables selection in discriminant analysis ## training sample : n = 1900, p = 14 variables data.learn <- scenarioCor[1:1900,1:14] labels.learn <-scenarioCor[1:1900,15] ## testing sample : n = 100, p = 14 variables data.test <- scenarioCor[1901:2000,1:14] labels.test <-scenarioCor[1901:2000,15] lambda <- seq(20, 50, length = 10) simulate.da <- SelvarLearnLasso(data.learn, labels.learn, lambda, rho, hybrid.size, models, regModel, indepModel, data.test, labels.test) ## End(Not run)
Want to suggest features or report bugs for rdrr.io? Use the GitHub issue tracker.