SEARCH RESULT

General principles for constructing a taxonomic identification system for big data catalogs are proposed. Several subject domains are researched, where a taxonomic description and identification traditionally used, are researched. On the fashionable clothes styles domain example, the use of taxonomic systematization and identification in domains non-traditional for this task is also considered. On basis of the results of the subject domains comparative analysis, which are similar in structure, a unified data model of a multi-level computer catalog of objects (taxa) and a taxonomic identification keys for it are proposed. This allowed to offer a unified approach for computer taxonomic catalogs implementation along with identification system. The general requirements for the taxonomic identification computer system are formulated, including the using of multi-input polytomous key; the hierarchical organization of taxa and taxonomic key characteristics; the possibility of step-by-step viewing and saving the selected characteristics; exclusion of conflicting key characteristics choice. The proposed approach has been tested in several subject domains, including bad formalized ones. The identification system usefulness is characterized by the efficiency of the searching and the convenience of constructing a taxonomic keys set. An approach to the constructed taxonomic key quality estimation is proposed based on using data mining methods. For some subject domains, completeness estimation of the taxonomic keys set is considered. A constructed hierarchical taxonomic key quality estimation analysis has been performed. A set of key features to determine their completeness and consistency is analyzed.

International Congress on 3D Printing (Additive Manufacturing) Technologies and Digital Industry
3D-PTC2019

Tatyana Petrushina Eugene Malakhov Natalia Trubina