Comparison Of Neural Network And Differential Evolution In Estimation Of Air Quality Using Mean Square Error

Softcomputing techniques are fast becoming reliable and efficient means of prediction and estimation. This has made their application more wide spread in recent years. With the growing need for intelligent devices and systems comes the need to explore these techniques even further. This paper applies neural networks and differential evolution (two of the most effective softcomputing algorithms) to the estimation of air quality and compares the accuracy of their results using the mean square error (MSE) method. Air pollution is an ever increasing menace in major cities around the world. Air contaminants such as those from motor vehicles and industrial wastes are the most common forms of pollutants. The health implications of inhaling contaminated air are evident in the growing number of cases of lung cancer and tuberculosis. Since these contaminants are invisible to the naked eye, it becomes necessary to implement an algorithm which can accurately identify them especially when their concentration becomes a threat to human health. The aim of this paper is to develop an effective algorithm to achieve this by comparing the efficacy of both neural networks and differential evolution in the determination of the concentration of air pollutants.The air component markers being analysed include oxides of carbon, nitrogen, sulphur and also ammonia. The study also intends to identify the most potent sources of air pollution by analysing air samples obtained at various locations within Kano city in Nigeria.