A Hybrid Model of a Flexible Rough Neural Network and Genetic Algorithm
(FRNN-GA) in Numerical Weather Forecasting Using Emotional Learning
Strategy
Abstract
The forecast of atmospheric processes is of great importance in planning
and management. Meteorological parameters are highly nonlinear
phenomena, varying with time and location, and many climatic factors
affect their changes. In this paper, a hybrid method consisting of a
flexible rough neural network (FRNN) and a genetic algorithm (GA)
proposed for forecasting meteorological parameters. Emotional learning
process used for learning rough neural network parameters by having
memories of previous learning history parameters. The forecasting
parameters used in this study are temperature, pressure, relative
humidity, wind speed, dew point, and visibility. In FRNN neurons,
instead of using an activating function, a combination of three
different sigmoid, tangent hyperbolic and linear functions is used to
add neuron flexibility. The genetic algorithm has also been used to
select the number and type of network input parameters. It expected that
the proposed method will work well for a chaotic system of uncertainty.
To evaluate the performance of the proposed hybrid method, data from the
Tehran Meteorological Database from 2008 to 2012 used. The results of
the implementation demonstrate the effective efficiency of FRNN-GA in
forecasting meteorological parameters compared to similar methods, and
using emotional learning increased the accuracy.