iteration {GIC}

A General Iterative Clustering Algorithm

Description

An algorithm that improves the proximity matrix (PM) from a random forest (RF) and the resulting clusters from an arbitrary cluster algorithm as measured by the silhouette score. The initial PM, that uses unlabeled data, is produced by one of many ways to provide psuedo labels for a RF. After running a cluster program on the resulting initial PM, cluster labels are obtained. These are used as labels with the same feature data to grow a new RF yielding an updated proximity matrix. This is entered into the clustering program and the process is repeated until convergence.

Usage

iteration(data,initiallabel,ntree=500)

Arguments

data

an input dataframe without label

initiallabel

a vector of label to begin with

ntree

the number of trees (default 500).

Details

This code requires initial labels as input, which can be obtained by any method of the users choice. As an alternative, Breimans' unsupervised method or Siegel and her colleagues' purposeful clustering method to obtain initial labels, use the function GIC

Value

An object of class iteration, which is a list with the following components:

PAM

output final PAM information

randomforest

output final randomforest information

clustering

A vector of integers indicating the cluster to which each point is allocated.

silhouette_score

A value of mean silhouette score for clusters

plot

A scatter plot which X-axis, y-axis, and color are first important feature, second important feature, and final clusters, respectively.

References

Breiman, L. (2001), Random Forests, Machine Learning 45(1), 5-32.

Siegel, C.E., Laska, E.M., Lin, Z., Xu, M., Abu-Amara, D., Jeffers, M.K., Qian, M., Milton, N., Flory, J.D., Hammamieh, R. and Daigle, B.J., (2021). Utilization of machine learning for identifying symptom severity military-related PTSD subtypes and their biological correlates. Translational psychiatry, 11(1), pp.1-12.

Examples

data(iris)
##Using KMEANS to find inital label
cl=kmeans(iris[,1:4],3)
###Doing GIC to find final clustering
rs=iteration(iris[,1:4],cl$cluster,ntree=100)
print(rs$clustering)