In recent years, the relevance of classifying information using neural networks has been growing. This is due to the fact that the amount of data that needs to be processed is getting larger every day. Almost all modern software systems are characterized by a large variety of interacting software modules, which, in turn, increases the complexity of data processing. At this stage, software developers use different methods. Neural networks allow you to achieve the highest possible speeds and high accuracy in working with a large amount of information compared to other methods of classification and data processing. In this regard, the task of developing classification methods is becoming more and more urgent.
The article considers an approach to classifying a sample of machine models described by a large number of features. Two of the most well-known types of neural networks are used as a classification tool, namely the multi-layer perceptron and the recognition network. Due to the fact that no high quality classification was obtained on any network type using a full set of features, the authors applied the principal component method (PCA) to reduce the feature space, which resulted in a significant increase in the quality of classification.
The developed approach can be used to classify machine models that are not represented in the sample. In addition, the article illustrates the fact that the choice of a classification method largely depends on the type of a subject area and the nature of the sample.
The software implementation is based on the MATLAB system, which provides a variety of tools and methods for preparing, analyzing, and visualizing data.
1. Gorban' A.N. Obuchenie neironnykh setei [Training of neural networks]. Moscow, Paragraph Publ., 1990. 160 p.(In Russian).
2. Gorban' A.N., Dunin-Barkovskii V.L. e. a. Neiroinformatika [Neuroinformatics]. Novosibirsk, Nauka Publ., 1998. 296 p.
3. Goodfellow Y., Benjio I., Courville A. Deep learning. Cambridge, The MIT Press, 2017 (Russ. ed.: Gudfellou Ya., Bendzhio I., Kurvill' A. Glubokoe obuchenie. Moscow, DMK-Press, 2017. 652 p.).
4. Callan R. The essence of neural networks. London, Prentice Hall Europe, 1999 (Russ. ed.: Kallan R. Osnovnye kontseptsii neironnykh setei. Moscow, Williams Publ., 2001. 288 p.).
5. Kruglov V.V., Borisov V.V. Iskusstvennye neironnye seti: teoriya i praktika [Artificial neural networks. Theory and practice]. Moscow, Hotline-Telecom, 2001. 382 p.
6. Goodfellow Y., Benjio I., Courville A. Deep learning. Cambridge, The MIT Press, 2016.
7. Merrel J.-P. Neural network. - Dordrecht: Springer, 2006. - P. 15–79.
8. Menezes A.J., Oorschot P.V., Vanstone S.A. Handbook of applied cryptography. Boca Raton, CRC Press, 1996. 816 p.
9. Terekhov V.A., Efimov D.V., Tyukin I.Yu. Neirosetevye sistemy upravleniya [Neural network control system]. Moscow, Vysshaya shkola Publ., 2002. 184 p.
Sosinskaya S.S., Rogacheva S.A. Klassifikatsiya stankov, opisyvaemykh bol'shim koliche-stvom priznakov, na osnove neironnoi seti raspoznavaniya i reduktsii prostranstva priznakov [Classification of machine tools described by a large number of features based on the neural network of recognition and attribute space reduction]. Nauchnyi vestnik Novosibirskogo gosudarstvennogo tekhnicheskogo universiteta = Science bulletin of the Novosibirsk state technical university, 2020, no. 1 (78), pp. 107–118. DOI: 10.17212/1814-1196-2020-1-107-118.