Scientific journal
Scientific Review. Technical science
ISSN 2500-0799
ПИ №ФС77-57440

Personal portfolio

ON THE CONSTRUCTION OF INFORMATION SYSTEM OF SELECTION OF THE OPTIMAL CHARACTERISTICS OF ARTIFICIAL NEURAL NETWORKS FOR FORECASTING PROBLEMS IN THE POWER SECTOR

Djachenko R.A. 1
1 Kuban State Technological University
The questions related to the creation of information support for the selection of the optimum characteristics of artificial neural networks. This object is achieved in the solution of a system of projections of energy facilities. The author has analyzed the extant software for forecasting in power. In the selected base models of information systems in forecasting power, described in the UML-activity diagrams, identified strengths and weaknesses, trends. The author further the task of selecting the optimum characteristics of artificial neural networks has been formalized in the form of a problem by minimizing training time at a fixed sample solution is for large training samples takes a long time. To solve the problem of long training by the author proposed a method based on the concept of distributed computing MapReduce.