Journal Article

A comparison of Artificial Neural Networks (ANN) and Local Linear Regression (LLR) techniques for predicting monthly reservoir levels

Hassan, Muhammad, Shamim, Muhammad Ali, Ahmad, Sameel, Zeeshan, Muhammad

2016

Overview

Storage dams play a very important role in irrigation especially during lean periods. For proper regulation one should make sure the availability of water according to needs and requirements. Normally regression techniques are used for the estimation of a reservoir level but this study was aimed to account for a non-linear change and variability of natural data by using Gamma Test, for input combination and data length selection, in conjunction with Artificial Neural Networking (ANN) and Local Linear Regression (LLR) based models for monthly reservoir level prediction. Results from both training and validation phase clearly indicate the usefulness of both ANN and LLR based prediction techniques for Water Management in general and reservoir level forecasting in particular, with LLR outperforming the ANN based model with relatively higher values of Nash-Sutcliffe model efficiency coefficnet (R 2 ) and lower values of Root Mean Squared Error (RMSE) and Mean Biased Error (MBE). The study also demonstrates how Gamma test can be effectively used to determine the ideal input combination for data driven model development.

• Publisher: Korean Society of Civil Engineers,Springer Nature B.V,대한토목학회
• Subject: Artificial neural networks / Civil Engineering / Engineering / Error analysis / Errors / Geotechnical Engineering & Applied Earth Sciences / Industrial Pollution Prevention / Irrigation / Learning theory / Mathematical models / Neural networks / Regression / Regression analysis / Regressions / Reservoirs / Root-mean-square errors / Water availability / Water Engineering / Water management / 토목공학