A Comparative Study of Artificial Neural Networks and Adaptive Nero-Fuzzy Inference System for Forecasting Daily Discharge of a Tigris River

摘要

Forecasting of the river discharge is important for convenient water resources management and planning specially in dry and warm region. This study prepares the application and comparison of artificial neural networks (ANNs) with back-propagation algorithm and adaptive nero-fuzzy inference system (ANFIS) for predicting the daily discharge in the Tigris River in Qurnah, Basrah, south of Iraq. The overall strategy in this study is to construct and develop three models for both ANNs and ANFIS techniques to forecast the daily discharge of the river depending on observed data was taken in earlier years. Three statistical parameters (root mean square error, efficiency coefficient and coefficient of correlation) are used to evaluate the performance of the three models in each technique. Depending on results of statistical results, both techniques have the ability to predict the river discharge. Also, it is found that the third model is better than the first and second models in both techniques to predict the daily discharge in the river. At last, in all models, statistical parameters and graphical results showed that the convergence between observed and predicted data is very good by using ANFIS models as compared to ANNs models.