A Method to Predict Heavy Precipitation using
the Artificial Neural Networks with an Application
Authors
Sulaiman Junaida, Hideo Hirose
Source
7th International Conference on Computing
and Convergence
Technology (ICCIT2012), pp.687-691, December
3-5, 2012, Seoul, Republic of Korea
Abstract
Many
flood occurrences are associated with heavy precipitation events.
By using Artificial Neural Network (ANN) in predicting heavy precipitation,
we can make projections of such events in the future. Since last
decade, ANN applications in hydrology have grown extensively. The
choice of input variables becomes crucial in identifying the optimal
ANN model. This paper describes the method to predict the heavy
precipitation by using ANN coupled with input variable selection
(IVS) method to identify the significant inputs for heavy precipitation
prediction in Malaysia. A stepwise regression method is used to
find significant inputs which influences the heavy precipitation
in the first phase. In the second phase, several models of ANN
are built using different input sets including those which are
selected during IVS process. The results of experiment revealed
that the stepwise regression method increases the prediction performance
of the ANN models such that they provide the useful information
on the relationship between the heavy precipitation and other potential
input variables
Key Words
Artificial
neural network; heavy precipitation events; stepwise regression;
climate indices