STUDY OF ARTIFICIAL INTELLIGENCE METHODS FOR CONSTRUCTING A CLASSIFICATION MODEL OF SUCCESSFUL TRANSACTION IMPLEMENTATION IN THE SS7 NETWORK

UDC 004.8
DOI:10.26102/2310-6018/2019.26.3.023

A.V. Roslyakov, S.V. Palmov, E.V. Glushak


Telecommunications have a decisive influence on the development of human society. The type of communication can be improved in various ways. Artificial intelligence allows to contribute to the solution of the abovementioned problem. However, there is the problem of choosing the method that is most suitable for solving a specific task in a particular subject area. This is due to the large number of existing artificial intelligence tools, as well as a significant variety of situations that require consideration of certain restrictions and nuances when analyzing them. The authors of the paper conduct an experiment that aimed to simplify the solution of the stated problem when it is necessary to build a classification model that determines the success of a SS7 network transaction implementation in the provision of voice and SMS services in a mobile communication network using transferred mobile subscriber numbers. The capabilities of the five methods are analyzed: decision tree, support vector machines, random forest, neural network and naive Bayes classifier. The classification models generated by the abovementioned methods test for compliance with two requirements: the reliable predictions generation and the results stability. Models quality evaluated by three metrics: F-measure, specificity and standard deviation. The experiment used real depersonalized statistics obtained in the network of a large mobile operator. After carrying out the relevant calculations and comparisons, it was found that the decision tree method usage seems to be the most preferable, since it forms the highest quality classification models.

Keywords: artificial intelligence, telecommunications, F-measure, decision tree, classification model, SS7.

Full text:
RoslyakovSoavtori_3_19_1.pdf