A Topological Clustering on Evolutionary Data

The objective of this paper is to propose a topological approach of clustering in evolutionary data analysis. We are interested in clustering resulting from exploratory methods of joint analysis of several data tables, methods applied more particularly to temporal data. The clustering is one of the most widely used approaches to exploring multidimensional data. The two common unsupervised clustering strategies are Hierarchical Ascending Clustering (HAC) and k-means partitioning used to identify groups of similar objects in a dataset to divide it into homogeneous groups. The proposed approach, called Topological Clustering on Evolutionary Data (TCED), is based on the notion of neighborhood graphs in an evolutionary data context. It makes it possible to simultaneously explore several tables of data collected at different times on the same individual-rows, the variables possibly being different according to the tables considered. The columns-variables of each table are more-or-less correlated or linked according to whether the variable type. It analyzes in each table the structure of the correlations or associations observed between the variables according to their quantitative, qualitative type or a mixture of both. The proposed TCED approach is presented and illustrated here using a real dataset with quantitative variables. Its results are compared with those resulting from the unsupervised clustering on evolutionary data analysis method—Multiple Factorial Analysis (MFA).