Experimental study on reduction of one-sided concept lattice based on subsets quality measure

The methods from the area of Formal Concept Analysis (or FCA) are usually used for conceptual-based exploratory data-mining tasks. One of the methods used in this area is based on one-sided fuzzification of standard approach, which provides the clusters of objects structured in concept hierarchy according to the specified set of fuzzy attributes. The limiting problem of FCA-based approaches is that they produce large set of concepts, which can be problematic for the interpretability of results and practical usage of the method. We have designed a method for evaluation of concepts from so-called generalized one-sided concept lattice, which is based on the quality measure of objects subsets. This method selects most relevant concepts and therefore leads to reduction of concept lattice. In this paper we will provide experimental study on the reduction, which can be achieved according to different settings of randomly generated data sets.