A Neural-Like Hierarchical Structure in the Problem of Automatic Generation of Hypotheses of Rules for Classifying the Objects Specified by Sets of Fuzzy Features

  • Konstantin V. SidorovEmail author
  • Natalya N. FilatovaEmail author
  • Pavel D. ShemaevEmail author
Conference paper
Part of the Advances in Intelligent Systems and Computing book series (AISC, volume 874)


The article describes a toolkit from the class of hybrid neural network models, which integrates a strategy of generalization and methods of fuzzy data processing. Evolution of properties of these models through the tasks of rules hypotheses construction for the purpose of classification of objects specified by fuzzy features can improve an efficiency of interpreters, designed for signals of bioengineering systems and for other data recorded from medical equipment. Considered the possibility of neural-like hierarchical structure (NLHS) utilization for the task of automatic generation of rules hypotheses for classification of objects determined by discrete features. We propose a hybrid algorithm to solve this problem, which acts as generalization generator, creating descriptions close in form to expert’s constructions. The algorithm includes in a composition of classification rules only the most significant features. The apparatus of fuzzy sets can be used for description of feature spaces. This article describes the software which implements the presented algorithm and also demonstrates results of its work with artificial data sets. The Fisher’s irises classification task was used to assess program capabilities. The variants of NLHS-based classification rules, created for a training set, considers as hypotheses, which verification is carried out using a test set and an expert analysis of their structure. Options of rules correction are checked with expert’s support besides the accuracy and completeness of the corrected rules are also evaluated. The constructed NLHS and corresponding training set is saved as a basic archive explaining formalisms contained in interpreter rules.


Algorithm Classification Neural-like hierarchical structure Fuzzy set Fuzzy sign Production rule Training set Test set 


  1. 1.
    Filatova, N.N., Dmitriev, G.A., Grigorieva, O.M.: Methods and algorithms for classifying graphic objects in problems of medical diagnostics. Tver State Technical University Publisher, Tver (2011). (in Russian, Metody i Algoritmy Klassifikatcii Graficheskikh Obektov v Zadachakh Meditcinskoi Diagnostiki)Google Scholar
  2. 2.
    Vagin, V.N., Golovina, E.Ju., Zagorjanskaja, A.A., Fomina, M.V.: Reliable and Plausible Conclusion in Intellectual Systems, 2nd edn. Fizmatlit Publisher, Moscow (2008). (in Russian, Dostovernyj i Pravdopodobnyj Vyvod v Intellektual’nyh Sistemah)Google Scholar
  3. 3.
    Yarushkina, N.G., Afanaseva, T.V., Perfileva, I.G.: Intelligent Time Series Analysis: Textbook. Ulyanovsk State Technical University Publisher, Ulyanovsk (2010). (in Russian, Intellektual’nyy Analiz Vremennykh Ryadov: Uchebnoe Posobie)Google Scholar
  4. 4.
    Loskutov, A.Yu.: Time Series Analysis: Lectures. Moscow State University Publisher, Moscow (2006). (in Russian, Analiz Vremennykh Ryadov: Kurs Lektsiy)Google Scholar
  5. 5.
    Ifeachor, E.C., Jervis, B.W.: Digital Signal Processing: A Practical Approach, 2nd edn. Pearson Education, London (2002)Google Scholar
  6. 6.
    Gladun, V.P.: Partnership with Computers. Man-Machine Purposeful Systems. Port-Royal Publisher, Kiev (2000). (in Russian, Partnerstvo s Kompiuterom. Cheloveko-Mashinnye Tceleustremlennye Sistemy)Google Scholar
  7. 7.
    Gladun, V.P.: The growing pyramidal networks. Artif. Intell. News 1, 30–40 (2004). (in Russian, Novosti Iskusstvennogo Intellekta)Google Scholar
  8. 8.
    Khaneev, D.M., Filatova, N.N.: The pyramidal network for classification of objects, presented by fuzzy features. Izvestiya SFedU. Eng. Sci. 134(9), 45–49 (2012). (in Russian, Izvestiya SFedU. Tekhnicheskie Nauki)Google Scholar
  9. 9.
    Filatova, N.N., Khaneev, D.M.: Use of neuro-like structures for automatic generation of hypotheses for classification rules. Fuzzy Syst. Soft Comput. 8(1), 27–44 (2013). (in Russian, Nechetkie Sistemy i Mjagkie Vychislenija)Google Scholar
  10. 10.
    Filatova, N.N., Khaneev, D.M., Sidorov, K.V.: Signals interpreter based on neural-like hierarchical structure. Softw. Syst. 105(1), 92–97 (2014). (in Russian, Programmnye Produkty i Sistemy)Google Scholar
  11. 11.
    Filatova, N.N., Khaneev, D.M., Sidorov, K.V.: Diagrams classification algorithm with consequent enlarging of features. Softw. Syst. 107(3), 78–86 (2014). (in Russian, Programmnye Produkty i Sistemy)CrossRefGoogle Scholar
  12. 12.
    Fisher, R.A.: The use of multiple measurements in taxonomic problems. Ann. Eugen. 7, 179–188 (1936)CrossRefGoogle Scholar
  13. 13.
    UCI Machine Learning Repository: Iris Data Set. Accessed 05 May 2018
  14. 14.
    Marcellus, D.H.: Expert Systems Programming in Turbo Prolog. Prentice Hall Inc., Upper Saddle River (1989)Google Scholar
  15. 15.
    Kolesnikov, A.V.: Gibrid Intelligent Systems: A Theory and Development Technology. St. Petersburg State Polytechnic University Publisher, St. Petersburg (2001). (in Russian, Gibridnye Intellektualnye Sistemy: Teoriya i Tekhnologiya Razrabotki)Google Scholar

Copyright information

© Springer Nature Switzerland AG 2019

Authors and Affiliations

  1. 1.Department of Information TechnologiesTver State Technical UniversityTverRussia

Personalised recommendations