| Title: | K-Means Clustering with Build-in Missing Data Imputation |
|---|---|
| Description: | This k-means algorithm is able to cluster data with missing values and as a by-product completes the data set. The implementation can deal with missing values in multiple variables and is computationally efficient since it iteratively uses the current cluster assignment to define a plausible distribution for missing value imputation. Weights are used to shrink early random draws for missing values (i.e., draws based on the cluster assignments after few iterations) towards the global mean of each feature. This shrinkage slowly fades out after a fixed number of iterations to reflect the increasing credibility of cluster assignments. See the vignette for details. |
| Authors: | Oliver Pfaffel |
| Maintainer: | Oliver Pfaffel <[email protected]> |
| License: | GPL-3 |
| Version: | 0.2.4 |
| Built: | 2025-03-11 04:01:32 UTC |
| Source: | https://github.com/o1iv3r/clustimpute |
Internal function of ClustImpute: check new centroids for duplicate rows and replace with random draws in this case.
check_replace_dups(centroids, X, seed)check_replace_dups(centroids, X, seed)
centroids |
Matrix of centroids |
X |
Underlying data matrix (without missings) |
seed |
Seed used for random sampling |
Returns centroids where duplicate rows are replaced by random draws
Clustering algorithm that produces a missing value imputation using on the go. The (local) imputation distribution is defined by the currently assigned cluster. The first draw is by random imputation.
ClustImpute( X, nr_cluster, nr_iter = 10, c_steps = 1, wf = default_wf, n_end = 10, seed_nr = 150519, assign_with_wf = TRUE, shrink_towards_global_mean = TRUE )ClustImpute( X, nr_cluster, nr_iter = 10, c_steps = 1, wf = default_wf, n_end = 10, seed_nr = 150519, assign_with_wf = TRUE, shrink_towards_global_mean = TRUE )
X |
Data frame with only numeric values or NAs |
nr_cluster |
Number of clusters |
nr_iter |
Iterations of procedure |
c_steps |
Number of clustering steps per iteration |
wf |
Weight function. Linear up to n_end by default. Used to shrink X towards zero or the global mean (default). See shrink_towards_global_mean |
n_end |
Steps until convergence of weight function to 1 |
seed_nr |
Number for set.seed() |
assign_with_wf |
Default is TRUE. If set to False, then the weight function is only applied in the centroid computation, but ignored in the cluster assignment. |
shrink_towards_global_mean |
By default TRUE. The weight matrix w is applied on the difference of X from the global mean m, i.e, (x-m)*w+m |
Completed data without NAs
For each row of complete_data, the associated cluster
For each cluster, the coordinates of the centroids in tidy format
For each cluster, the coordinates of the centroids in matrix format
Mean of the imputed variables per draw
Standard deviation of the imputed variables per draw
# Random Dataset set.seed(739) n <- 750 # numer of points nr_other_vars <- 2 mat <- matrix(rnorm(nr_other_vars*n),n,nr_other_vars) me<-4 # mean x <- c(rnorm(n/3,me/2,1),rnorm(2*n/3,-me/2,1)) y <- c(rnorm(n/3,0,1),rnorm(n/3,me,1),rnorm(n/3,-me,1)) dat <- cbind(mat,x,y) dat<- as.data.frame(scale(dat)) # scaling # Create NAs dat_with_miss <- miss_sim(dat,p=.1,seed_nr=120) # Run ClustImpute res <- ClustImpute(dat_with_miss,nr_cluster=3) # Plot complete data set and cluster assignment ggplot2::ggplot(res$complete_data,ggplot2::aes(x,y,color=factor(res$clusters))) + ggplot2::geom_point() # View centroids res$centroids# Random Dataset set.seed(739) n <- 750 # numer of points nr_other_vars <- 2 mat <- matrix(rnorm(nr_other_vars*n),n,nr_other_vars) me<-4 # mean x <- c(rnorm(n/3,me/2,1),rnorm(2*n/3,-me/2,1)) y <- c(rnorm(n/3,0,1),rnorm(n/3,me,1),rnorm(n/3,-me,1)) dat <- cbind(mat,x,y) dat<- as.data.frame(scale(dat)) # scaling # Create NAs dat_with_miss <- miss_sim(dat,p=.1,seed_nr=120) # Run ClustImpute res <- ClustImpute(dat_with_miss,nr_cluster=3) # Plot complete data set and cluster assignment ggplot2::ggplot(res$complete_data,ggplot2::aes(x,y,color=factor(res$clusters))) + ggplot2::geom_point() # View centroids res$centroids
Default weight function. One minus the return value is multiplied with missing(=imputed) values. It starts with 1 and goes to 0 at n_end.
default_wf(n, n_end = 10)default_wf(n, n_end = 10)
n |
current step |
n_end |
steps until convergence of weight function to 0 |
value between 0 and 1
x <- 0:20 plot(x,1-default_wf(x))x <- 0:20 plot(x,1-default_wf(x))
Simulates missing at random using a normal copula to create correlations between the missing (type="MAR"). Missings appear in each column of the provided data frame with the same ratio.
miss_sim(dat, p = 0.2, type = "MAR", seed_nr = 123)miss_sim(dat, p = 0.2, type = "MAR", seed_nr = 123)
dat |
Data frame with only numeric values |
p |
Fraction of missings (for entire data frame) |
type |
Type of missingness. Either MCAR (=missing completely at random) or MAR (=missing at random) |
seed_nr |
Number for set.seed() |
data frame with only numeric values and NAs
data(cars) cars_with_missings <- miss_sim(cars,p = .2,seed_nr = 4) summary(cars_with_missings)data(cars) cars_with_missings <- miss_sim(cars,p = .2,seed_nr = 4) summary(cars_with_missings)
Returns a plot with the marginal distributions by cluster and feature. The plot shows histograms or boxplots and , as a ggplot object, can be modified further.
## S3 method for class 'kmeans_ClustImpute' plot( x, type = "hist", vline = "centroids", hist_bins = 30, color_bins = "#56B4E9", color_vline = "#E69F00", size_vline = 2, ... )## S3 method for class 'kmeans_ClustImpute' plot( x, type = "hist", vline = "centroids", hist_bins = 30, color_bins = "#56B4E9", color_vline = "#E69F00", size_vline = 2, ... )
x |
an object returned from ClustImpute |
type |
either "hist" to plot a histogram or "box" for a boxplot |
vline |
for "hist" a vertical line is plotted showing either the centroid value or the mean of all data points grouped by cluster and feature |
hist_bins |
number of bins for histogram |
color_bins |
color for the histogram bins |
color_vline |
color for the vertical line |
size_vline |
size of the vertical line |
... |
currently unused |
Returns a ggplot object
Prediction method
## S3 method for class 'kmeans_ClustImpute' predict(object, newdata, ...)## S3 method for class 'kmeans_ClustImpute' predict(object, newdata, ...)
object |
Object of class kmeans_ClustImpute |
newdata |
Data frame |
... |
additional arguments affecting the predictions produced - not currently used |
integer value (cluster assignment)
# Random Dataset set.seed(739) n <- 750 # numer of points nr_other_vars <- 2 mat <- matrix(rnorm(nr_other_vars*n),n,nr_other_vars) me<-4 # mean x <- c(rnorm(n/3,me/2,1),rnorm(2*n/3,-me/2,1)) y <- c(rnorm(n/3,0,1),rnorm(n/3,me,1),rnorm(n/3,-me,1)) dat <- cbind(mat,x,y) dat<- as.data.frame(scale(dat)) # scaling # Create NAs dat_with_miss <- miss_sim(dat,p=.1,seed_nr=120) res <- ClustImpute(dat_with_miss,nr_cluster=3) predict(res,newdata=dat[1,])# Random Dataset set.seed(739) n <- 750 # numer of points nr_other_vars <- 2 mat <- matrix(rnorm(nr_other_vars*n),n,nr_other_vars) me<-4 # mean x <- c(rnorm(n/3,me/2,1),rnorm(2*n/3,-me/2,1)) y <- c(rnorm(n/3,0,1),rnorm(n/3,me,1),rnorm(n/3,-me,1)) dat <- cbind(mat,x,y) dat<- as.data.frame(scale(dat)) # scaling # Create NAs dat_with_miss <- miss_sim(dat,p=.1,seed_nr=120) res <- ClustImpute(dat_with_miss,nr_cluster=3) predict(res,newdata=dat[1,])
Returns a plot with the marginal distributions by cluster and feature. The plot shows histograms or boxplots and , as a ggplot object, can be modified further.
## S3 method for class 'kmeans_ClustImpute' print(x, ...)## S3 method for class 'kmeans_ClustImpute' print(x, ...)
x |
an object returned from ClustImpute |
... |
currently unused |
No return value (print function)
Computes one minus the ratio of the sum of all within cluster variances by the overall variance
var_reduction(clusterObj)var_reduction(clusterObj)
clusterObj |
Object of class kmeans_ClustImpute |
integer value typically between 0 and 1
# Random Dataset set.seed(739) n <- 750 # numer of points nr_other_vars <- 2 mat <- matrix(rnorm(nr_other_vars*n),n,nr_other_vars) me<-4 # mean x <- c(rnorm(n/3,me/2,1),rnorm(2*n/3,-me/2,1)) y <- c(rnorm(n/3,0,1),rnorm(n/3,me,1),rnorm(n/3,-me,1)) dat <- cbind(mat,x,y) dat<- as.data.frame(scale(dat)) # scaling # Create NAs dat_with_miss <- miss_sim(dat,p=.1,seed_nr=120) res <- ClustImpute(dat_with_miss,nr_cluster=3) var_reduction(res)# Random Dataset set.seed(739) n <- 750 # numer of points nr_other_vars <- 2 mat <- matrix(rnorm(nr_other_vars*n),n,nr_other_vars) me<-4 # mean x <- c(rnorm(n/3,me/2,1),rnorm(2*n/3,-me/2,1)) y <- c(rnorm(n/3,0,1),rnorm(n/3,me,1),rnorm(n/3,-me,1)) dat <- cbind(mat,x,y) dat<- as.data.frame(scale(dat)) # scaling # Create NAs dat_with_miss <- miss_sim(dat,p=.1,seed_nr=120) res <- ClustImpute(dat_with_miss,nr_cluster=3) var_reduction(res)