acca: Average correlation clustering algorithm

In meantrix/corrP: Compute Correlations Type Analysis in Parallel

Average correlation clustering algorithm

Description

A C++ implementation of the Average Correlation Clustering Algorithm (ACCA) https://www.sciencedirect.com/science/article/pii/S1532046410000158, originally developed for genetic studies using Pearson correlation as a similarity measure. Unlike traditional clustering methods that rely on distance metrics such as Euclidean or Mahalanobis distance, ACCA groups data based on correlation patterns.
This implementation works directly with the correlation matrix derived from the corr_matrix function and supports mixed data types along with various correlation methods.
ACCA is an unsupervised clustering method, meaning it identifies patterns without predefined labels. Similar to k-means, it requires defining the K parameter, which controls the number of clusters.

Usage

acca(m, k, ...)

## S3 method for class 'cmatrix'
acca(m, k, maxrep = 2L, maxiter = 100L, ...)

## S3 method for class 'matrix'
acca(m, k, maxrep = 2L, maxiter = 100L, ...)

Arguments

m	[matrix] correlation matrix from corr_matrix or a distance matrix.
k	[integer(1)] number of clusters considered.
...	Not used. Included for S3 method consistency.
maxrep	[integer(1)] Maximum number of consecutive iterations without changes in the cluster structure before stopping the algorithm. If the clusters remain identical for maxrep iterations in a row, the algorithm considers that it has converged and stops early, without reaching maxiter. It serves to speed up execution by detecting when the clustering has stabilized.
maxiter	[integer(1)] Maximum number of iterations allowed for the algorithm. Regardless of cluster stability, the algorithm will be forcibly stopped after maxiter iterations to prevent infinite loops or excessive computation. It acts as a hard upper limit on the runtime.

Value

[acca_list]
A list with the final result of the clustering method. That is, every element of the list group names of the variables belonging to each cluster k.

Author(s)

Igor D.S. Siciliani, Paulo H. dos Santos

References

Bhattacharya, Anindya, and Rajat K. De. "Average correlation clustering algorithm (ACCA) for grouping of co-regulated genes with similar pattern of variation in their expression values." Journal of Biomedical Informatics 43.4 (2010): 560-568.

Examples

# Clustering a correlation matrix with 3 clusters
x <- corrp::corrp(iris)
m <- corrp::corr_matrix(x)
result <- corrp::acca(m, k = 3)
print(result)

# Clustering with 5 clusters and increasing the maximum number of interactions
x <- corrp::corrp(iris)
m <- corrp::corr_matrix(x)
result <- corrp::acca(m, k = 5, maxiter = 200)
print(result)


# Adjusting the maximum number of iterations without change in clusters
x <- corrp::corrp(iris)
m <- corrp::corr_matrix(x)
result <- corrp::acca(m, k = 2, maxrep = 50)
print(result)

meantrix/corrP documentation built on June 12, 2025, 5:33 p.m.