A pyramidal classification for interval data using Galois lattice reduction
It is assumed that we observed for each of the \(n\) individuals \(u\) (elements, objects) of a set \(E=\{1,\dots,n\}\) a vector \(Y=(Y_1,\dots,Y_p)'\) of \(p\) quantitative interval-type variables \(Y_j\), all with the same domain \({\mathcal Y}_j=\mathbb{R}\). These observations are stored in a symbolic data array \(\underline{X}=(\xi_{uj})_{n\times p}\), where \(\xi_{uj}=Y_j(u)=[\underline{a}_{uj},\overline{a}_{uj}]\) is the interval from \({\mathcal Y}_j=\mathbb{R}\) which was observed for the \(j\)-th variable \(Y_j\) for the individual \(u\). The author shows how Galois lattices can be used for analyzing and interpreting symbolic data arrays in the case of interval variables. The classical theory of Galois lattices and the corresponding construction algorithms have been extended from the classical binary data tables to interval data (even with missing values) along the lines of the symbolic object formalism. The definition and some properties of Galois lattices are presented. Unfortunately, whereas the Galois lattice of all 'complete symbolic objects' can in principle be calculated and visualized graphically, this lattice is often too large and too complex to be easily displayed or interpreted. Therefore, an algorithm is proposed for reducing the resulting lattice. A pyramidal classification of the elements (complete assertion objects) of the lattice by means of a suitable dissimilarity coefficient is constructed. The dissimilarity coefficient is a heuristic one which combines numerical criteria with the lattice properties. Thereby, a generalized pyramid of assertion objects is obtained. For the entire collection see [Zbl 1039.62501].
This page was built for publication: A pyramidal classification for interval data using Galois lattice reduction
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2756652)
