Modeling of socio-economic systems on the basis of selective neurons network ensembles
Russian Plekhanov University, Educational and research center "Cybernetics" Department of Applied Informatics and Information Security, Tel. 89166019804, E-mail: Mikrukov.email@example.com
At present, neural network approaches, in particular, the use of ensembles of neural networks, which are an example of a collective solution of problems, are widely used to solve the problems of modeling socio-economic systems that belong to the class of hard-formalizable and weakly structured systems.
The use of ensembles of neural networks, in which the formation and training of a finite set of neural networks, the results of solving which are taken into account in the general solution, is expected to significantly improve the quality of the solution of a specific problem (data mining, forecasting, pattern recognition, classification, etc.) [1,2].
One of the fundamental problems of improving the process of modeling the functioning of an ensemble of neural networks from the point of view of increasing their accuracy and reliability is the generation of a variety of ensemble (differences of individual models). Aggregation of similar models in the ensemble can not lead to a significant improvement in the quality of the solution of the problem.
To resolve this contradiction, approaches have been developed for building a collective of electoral neural networks based on a new class of artificial neurons, so-called selective neurons, which differ from classical neurons in a more efficient way of processing input information that is close to a biological neuron .
The conducted researches showed that application of selective neural networks for building models of neural network ensembles allows to significantly increase the accuracy and reliability of simulation results.