v1.8.0

Automatic Trimmed & Sparse Clustering implementation (ATSC) and data cleaning methods.

Example:

library("evaluomeR")
seed = 13606
k.range = c(3,10)
cbi = "clara"
alpha = NULL
max_alpha = 0.1

data("golub")
golub = golub
colnames(golub)[colnames(golub) == 'Case'] <- 'Description'
head(golub)

golub_clean = evaluomeR::cleanDataset(golub, correlation_threshold=1)
pca_suitability = evaluomeR::PCASuitability(golub_clean$R, sig_level = 0.05)
dataset = golub_clean$dataset

if (pca_suitability$pca_suitable) {
    r_pca = evaluomeR::performPCA(dataset = dataset)
    dataset = r_pca$dataset_ncp
    evaluomeR::plotPCA_fviz_screeplot(r_pca$pca)
    evaluomeR::plotPCA_fviz_biplot(r_pca$pca)
}

r_atsc = evaluomeR::ATSC(data=dataset, k.range=k.range, cbi=cbi, alpha=alpha, 
                         max_alpha=max_alpha, gold_standard=gold_standard_vector, seed=seed)

print(paste0("Optimal k before ATSC: ", r_atsc$optimalK))
print(paste0("Optimal k with ATSC: ", r_atsc$optimalK_ATSC))

v1.7.9

Whole metrics analysis.

Optimal k analysis example with golub dataset:

library("evaluomeR") 
library("cancerclass")

load("leukemia.RData")
golub = as.data.frame(leukemia)
golub["Class"] = NULL
golub["sample"] = NULL
golub["type"] = NULL
golub["FAB"] = NULL
golub["gender"] = NULL

colnames(golub)[colnames(golub) == 'Case'] <- 'Description'

seed = 13606
k.range=c(3,10)
cbi = "clara"
stab_range = stabilityRange(data=golub, k.range=k.range, 
                            bs=100, seed=seed,
                            all_metrics=TRUE,
                            cbi=cbi)
stab = standardizeStabilityData(stab_range)

# Qual
qual_range = qualityRange(data=golub, k.range=k.range, 
                            all_metrics=TRUE, seed=seed,
                            cbi=cbi)

qual = standardizeQualityData(qual_range)
# K opt
k_opt = getOptimalKValue(stab_range, qual_range, k.range= k.range)
optimal_k = k_opt$Global_optimal_k
optimal_k_str = paste0("k_", optimal_k)
print(paste0("Optimal k: ", optimal_k))

v1.3.50

• Clusterboot interfaces can be set through 'cbi' parameter to stability methods.
  It takes one the following values: "kmeans", "clara", "clara_pam", "hclust", "pamk", "pamk_pam".
  For instance:

library(evaluomeR)
data("rnaMetrics")
stability(data=rnaMetrics, cbi="clara_pam", k=2, bs=100, getImages = FALSE)
stability(data=rnaMetrics, cbi="hclust", k=2, bs=100, getImages = FALSE)
stability(data=rnaMetrics, cbi="pamk", k=2, bs=100, getImages = FALSE)

v1.3.4

• Added plotMetricsClusterComparison method.

v1.3.3

• Optimized getOptimalKValue.
• Added two new methods: stabilitySet and qualitySet to process a list of discrete K values.
• Bug fixes.