Abstract:
Forecasting of weather is very popular in
nowadays. But forecasting the future from the
observed past is very difficult. There are several
forecasting methods for weather data. Among
them, evolving artificial neural networks are
suitable for weather time series forecasting
because of their abilities to learn and adapt a
new situation by recognizing new patterns in
previous data. However, ANNs present some
drawbacks such as over fitting and long time
processing. Using ANNs together with genetic
algorithm comes to solutions of these problems.
Genetic artificial neural networks (GANNs) can
give optimal forecasting result from the observed
past. In order to provide more effective ANNs,
the proposed system use cascade back
propagation instead of back propagation
method. Weather parameters (attributes) such as
rain fall (precipitation), humidity, wind force,
dew point, sea level, wind direction will also be
used to forecast maximum and minimum
temperature.
