|
Dr Fulufhelo Nelwamondo (left) and Ishmael Msiza
|
In a first-of-its-kind study in South Africa into water demand forecasting using computational intelligence techniques, CSIR researchers Dr Fulufhelo Nelwamondo and Ishmael Msiza found that the modelling technique based on how the human brain performs a particular task can be very useful. They determined that an artificial neural network (ANN) 'outsmarts' a support vector machine (SVM), which is a classifier derived from statistical learning theory.
Msiza says, "Modelling of water resource variables is a very broad field that includes modelling of water quality, the water demand trajectory and water reticulation networks, to mention but a few. There is currently a large pool of modelling techniques and hence there is always a need to investigate which technique is the most efficient for a particular application. Dr Nelwamondo and I decided to focus on the modelling of water demand in Gauteng because it is South Africa's industrial powerhouse; houses almost a quarter of the South African population; and consumes about 86% of the total water supply provided by Rand Water. We put ANNs against SVMs and the former proved to be 'a genius' when it comes to water demand forecasting."
Nelwamondo explains, "A neural network is an information processing paradigm inspired by the way biological nervous systems like the human brain process information. It is an exceptionally powerful instrument that has found successful use in mechanical, civil, aerospace and biomedical engineering; as well as finance. Due to their ability to gain meaning from complicated data, neural networks are employed to extract patterns and detect trends that are too complex to be noticed by many other computer techniques. A trained neural network can be considered as an expert in the category of information it has been provided to analyse. This 'expert' can then be used to provide predictions when presented with new situations."
Over the years, scientists have conducted various studies into the efficacy of ANNs against SVMs in diverse areas ranging from wastewater treatment processes forecasting, prediction of wind speed to financial time series forecasting. Given that in different situations certain techniques prove more efficient, studies must be conducted to ascertain which technique is most suitable for the application at hand.
Nelwamondo says that water scarcity in South Africa is serious. "Indicative of water shortages in certain parts of the country," Msiza says, "is the fact that we have programmes such as the Tugela-Vaal Water Transfer Scheme and the Lesotho Highlands Water Project."
Nelwamondo adds, "The CSIR's modelling and digital science research is directed at contributing to knowledge about physics and mathematics as embedded in both natural and man-made systems; engineering; and the rapidly advancing ICT infrastructure, aimed at improving quality of life. There is an urgent need for the development of tools that will assist in the effective management of water resources in South Africa and computational intelligence techniques have a significant role to play. We hope that these studies will assist in South Africa's quest to find efficient methods of monitoring and addressing our water challenges."
Their findings were published in the November 2008 issue of the Journal of Computers, an international scientific journal.
Enquiries: CSIR Communication
|
