Modeling time series data by using neural networks and genetic algorithms Conference

Tansel, IN, Yang, SY, Venkataraman, G et al. (1999). Modeling time series data by using neural networks and genetic algorithms . 9 1055-1060.

cited authors

  • Tansel, IN; Yang, SY; Venkataraman, G; Sasirathsiri, A; Bao, WY; Mahendrakar, N

abstract

  • Time series data with one output and data with one-input-one output were considered. Modeling accuracy, convenience and computational time of the following three approaches were studied: linear optimization, neural networks, and genetic algorithms. Linear optimization methods gave the best estimations. Genetic algorithm estimated the same values for the parameters of the models with the linear optimization if the boundaries of the parameters and the resolution are selected properly. It was possible to add non-linear characteristics to the model when the genetic algorithms were used. Neural networks made the worst estimations; however, they automatically include the non-linear characteristics into the model and require minimal theoretical background to work with.

publication date

  • December 1, 1999

start page

  • 1055

end page

  • 1060

volume

  • 9