Clustering the Constitutive Elements of Measuring Tables Data Structure

Abstract

This article deals with the clustering of the elements of a structure of juxtaposition of data measuring tables. One of the main issues in such problems is the selection of a one-dimensional quantity to represent the information included in the repeated observations of each variable. We propose the use of three different indices to measure the distance between elements of a structure and use the last one based on the Hilbert–Schmidt inner product for clustering purposes through an algorithmic procedure. The proposed algorithm is applied for clustering the customers of an electric company where each customer is described by a curve of load.

Keywords:

• Clustering criterion
• Constitutive element
• Factorial analysis
• Hilbert–Schmidt inner product
• Kullback–Leibler distance
• Similarity
• Structure of multiple table

Mathematics Subject Classification:

• 62B10
• 94A17
• 62H10

Notes

Value of the criterion for this partition Cr = 9.9455.

Value of criterion for this partition Cr = 4.0498.