GitHub - unina-sfere/sasfunclust: R package that implements the sparse and smooth functional clustering (SaS-Funclust) method. (original) (raw)

sasfunclust

The package sasfunclust implements the sparse and smooth functional clustering (SaS-Funclust) method proposed by Centofanti et al. (2021). SaS-Funclust is a new method for clustering functional data that aims to classify a sample of curves into homogeneous groups while jointly detecting the most informative portions of domain. The method relies on a general functional Gaussian mixture model whose parameters are estimated by maximizing a log-likelihood function penalized with the functional adaptive pairwise fusion penalty (FAPFP) and a roughness penalty. The package comprises two main functions sasfclust andsasfclust_cv. The former performs the SaS-Funclust for fixed number of clusters G, tuning parameter of the smoothness penalty_λ_ s, and tuning parameter of the FAPFP λ l. The latter executes the K-fold cross-validation procedure described in Centofanti et al. (2021) to choose G, λ s, and_λ_ l.

Installation

The released version of sasfunclust on CRAN can be installed with:

install.packages("sasfunclust")

The development version can be installed fromGitHub with:

install.packages("devtools")

devtools::install_github("unina-sfere/sasfunclust")

Example

This is a basic example which shows you how to apply the two main functions sasfclust and sasfclust_cv on a synthetic dataset generated as described in the simulation study of Centofanti et al. (2021).

We start by loading and attaching the sasfunclust package.

Then, we generate the synthetic dataset as follows.

n_i=20 train<-simulate_data("Scenario I",n_i=n_i,var_e = 1,var_b = 0.5^2)

To apply sasfclust_cv, sequences of G, λ s, and_λ_ l should be defined.

lambda_s_seq=10^seq(-4,-3) lambda_l_seq=10^seq(0,3) G_seq=2

And, then, sasfclust_cv is executed.

mod_cv<-sasfclust_cv(X=train$X,grid=train$grid,G_seq=G_seq, lambda_l_seq = lambda_l_seq,lambda_s_seq =lambda_s_seq,maxit = 50,K_fold = 5,q=30,ncores = 7)

The results are plotted.

By using the model selection method described in Centofanti et al. (2021), the optimal values of G, λ s, and λ l, are 2, 10 − 4, and 102, respectively.

Finally, sasfclust is applied with G, λ s, and_λ_ l fixed to their optimal values.

mod<-sasfclust(X=train$X,grid=train$grid,G=mod_cv$G_opt, lambda_l = mod_cv$lambda_l_opt,lambda_s =mod_cv$lambda_s_opt,maxit = 50,q=30)

The cluster membership vector and the plot of the estimated cluster mean functions and the classified curves are obtained as follows.

print(mod$clus$classes) plot(mod)

References

• Centofanti, F., Lepore, A., & Palumbo, B. (2021). Sparse and Smooth Functional Data Clustering. arXiv preprint arXiv:2103.15224.