Artificial neural networks for water level prediction based on Z-score technique in Kelantan river / Khairah Jaafar ...[et al.]

—In Malaysia, flood can happens annually anytime
of the year in multitude of ways. This study aimed to predict
water level at Jeti Kastam station (S6) in Kelantan River
using an Artificial Neural Networks (ANN) as a modelling
tool and validate the accuracy of the model. The Z-Score
technique is applied to previous rainfall and water level data
to all 6 stations along Kelantan River in identified the
significant stations before the successful data resulted will
fed to ANN model network. The ANN model was
formulated to simulate water level using feedforward
algorithm. Readings from 6 stations from rainfall stations
showed that S1, S2, S3 and S6 code station while S1 and S2
for water level station were significant value based on ZScore processing method. Total of 1095 data per station
collected from January 2013 until December 2015 was used
for training, validation and testing of the network model.
Mean Square Error (MSE) and Regression analysis, R are
calculated every node. The result showed that the 5 hidden
nodes in hidden layer revealed that the regression, R for
training, validation and testing were 0.9993, 0.9640 and
0.9989 respectively with MSE value was 2.14e-05. The result
of prediction model has found to be suitable to predict flood
model by training function feedforward optimization.