Application of Predictive Intelligence in Water Quality Forecasting of the River Ganga Using Support Vector Machines

Application of Predictive Intelligence in Water Quality Forecasting of the River Ganga Using Support Vector Machines

Anil Kumar Bisht, Ravendra Singh, Rakesh Bhutiani, Ashutosh Bhatt
Copyright: © 2019 |Pages: 13
DOI: 10.4018/978-1-5225-6210-8.ch009
OnDemand:
(Individual Chapters)
Available
$37.50
Current Special Offers
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

Predicting the water quality of rivers has attracted a lot of researchers all around the globe. A precise prediction of river water quality may benefit the water management bodies. However, due to the complex relationship existing among various factors, the prediction is a challenging job. Here, the authors attempted to develop a model for forecasting or predicting the water quality of the river Ganga using application of predictive intelligence based on machine learning approach called support vector machine (SVM). The monthly data sets of five water quality parameters from 2001 to 2015 were taken from five sampling stations from Devprayag to Roorkee in the Uttarakhand state of India. The experiments are conducted in Python 2.7.13 language (Anaconda2 4.3.1) using the radial basis function (RBF) as a kernel for developing the non-linear SVM-based classifier as a model for water quality prediction. The results indicated a prediction performance of 96.66% for best parameter combination which proved the significance of predictive intelligence in water quality forecasting.
Chapter Preview
Top

Background

During the past years the quality of water is deteriorating day-by-day that resulted in serious problem of water pollution. Therefore this issue of water security attracted a lot of attention of the researchers and academicians all around the world. A variety of research work in this fields has been going on using different techniques. Because of much difficult and highly non-linear nature of environmental systems the deterministic models which have been constructed so far are not up to the mark (Sarkar & Pandey, 2015). In addition to this, limited availability of water data and high cost in monitoring are main drawbacks associated with the process based modelling methodologies (Palani, 2008). Now-a-days various machine learning techniques are common in practice which use the given data in order to derive a solution for the given problem. Most of the water researches which were intended for water quantity modelling used the concept of artificial neural networks (ANNs). The ANNs have become tremendously popular and being used in many fields especially as an ecological modelling tool to predict/forecast water resource variables. The models based on ANN method are data-driven, nonlinear, flexible, require minimal human involvement and no detailed information regarding the system under concern (Yan, 2012). ANN based water quality models are faster, lucrative and requisite a low input data as compared with the process based models (Palani, 2008). Consequently, the modelling methods based on data driven principle are more in demand.

Complete Chapter List

Search this Book:
Reset