K-means Clustering in R Libraries {cluster} and {factoextra} for Grouping Oceanographic Data
Cluster analysis by k-means algorithm by R programming applied for the geological data analysis is the scope of the presented paper. The research object is the Mariana Trench, a hadal trench located in west Pacific Ocean. The study evaluates the similarity of the geological data by the analysis of their attributes. The original observation data set contained samples varying in parameters: geology (sediment thickness), tectonics (locations on the tectonic plates), volcanism (igneous volcanic areas), bathymetry (depth ranges) and geomorphology (slope steepness and aspect). The data pool was divided to the clusters using k-means algorithm with aim to detect similarities. Clustering was chosen as a main statistical method, since it enables detecting similar groups within the original data set by unsupervised classification. Technically, the research was performed using R language and its statistical libraries. The main R libraries include {cluster}, {factoextra}; minor libraries include {ggplot2}, {FactoMiner}, {openxlsx}, {carData}, {rio}, {car} and {flashClust}. Several clusters were tested from two to seven, the optimal number is defined as five. The results show visualized computations:
correlation matrix of the factors; comparison of the bi-factors showing pairwise correlation; pairwise comparative analysis showing influence of the variables as bi-factors: sediment thickness correlating with slope an- gles; correlation of the volcanic igneous areas with slope angles and as- pect degree. Four variables affect geomorphology: slope angle, sediment thickness, aspect degree, bathymetry and volcanism. The paper includes listings of R programming codes for repeatability of the algorithms in similar research.