Experimental study on time complexity of GOSCL algorithm for sparse data tables

In this paper we provide experimental study on time complexity of GOSCL algorithm according to the sparseness of the input data table. GOSCL is incremental algorithm for the creation of Generalized One-Sided Concept Lattices, which is related to well-known Formal Concept Analysis area, but with the possibility to work with different types of attributes and to produce one-sided concept lattice from the generalized one-sided formal context. Generally, FCA-based algorithms are exponential. However, in practice there are many inputs for which the complexity is reduced. One of the special cases is related to the high number of "zeros" (bottom elements) in data table for so-called sparse data matrices, which is characteristic for some inputs like document-term matrix in text-mining analysis. We describe experimentally the influence of sparseness of data tables on time complexity of GOSCL with different distributions of zeros generated artificially randomly or according to the standard text-mining datasets.