Experience of biclusterization application to data of agricultural crops sorts

The article is devoted to methods for analyzing object-attribute data in low structured areas. Specifically, a demonstrational example of biclusterization of similar data in the field of agricultural crops selection is considered. It is shown what opportunities are appearing for the systematization of empirical data. The data biclusterization method under consideration - a classical formal concept analysis - is an applied branch of lattice theory; a brief reference is given on its mathematical foundations. The source material for the demonstrational example was borrowed from the catalog of innovative developments of the Samara Research Institute of Agricultural Sciences named after N.M. Tulaykov. Sorts of soft spring wheat were chosen as objects, and several properties which are describing them were chosen as attributes. A number of these properties are scaled in accordance with state standard 9353-2016 for wheat. Based on the selected and adjusted material, the initial table “objects-properties”, or a formal context for the formal concept analysis, is formed. To process this data, we used the free software package ConExp, which implements a methodical complex for formal concept analysis. The main results that these methods give for selection data structuring are demonstrated. These results include a lattice of formal concepts, implications and associative rules on a set of attributes. Moreover, the result of the OntoWorker software package being developed at ICCS RAS - SamSC RAS is demonstrated. The outcome consists in transforming and reducing the lattice of formal concepts into a special taxonomy of formal concepts (classes) that is more convenient for the user to perceive and interpret. It is noted that advanced methods of formal concept analysis can take into account the incompleteness of the considered data, determined by the series multiplicity of measurement experiments with different degrees of results reliability, and by the presence of competing measurement procedures with different degrees of confidence in their results. This reflects the realities of the empirical information accumulation directly during multidimensional observations and measurements in selection probations. It was emphasized that the effective application of the considered methods is possible only in cooperation of knowledge data expert and specialists in the field of formal concept analysis.