• Corpus ID: 109456510

Разработка и апробация эмулятора нейросетевого моделирования для целей прогнозирования временных рядов

@inproceedings{2012,
  title={Разработка и апробация эмулятора нейросетевого моделирования для целей прогнозирования временных рядов},
  author={Д. В. Дмитриев and Д. А. Ляхманов and Э. С. Соколова},
  year={2012}
}

• Д. В. Дмитриев, Д. А. Ляхманов, Э. С. Соколова
• Published 2012
• Computer Science

We introduced the solution of the actual problem of automating design architecture of neural networks for forecasting, followed by appraisal at the time series of different nature. We took into account the following characteristics of neural networks in the design phase and training: the number of network layers, each layer of the number of neurons, the type of activation function, learning rate, the terms of graduation. Based on the results obtained from neural network modeling, provide recommendations for handling input to the forecast. The possibilities of accurate prediction of future performance of financial markets, including domestic, non-obvious factors and the influence of the external environment. We introduced the results of the emulator on forecasting financial time series of stock price in the long term are at the end.