Estimating Sectoral Pollution Load in Lagos, Nigeria Using Data Mining Techniques

TL;DR

This paper demonstrates that Artificial Neural Networks, especially the Time Lagged Recurrent Network, effectively estimate sectoral pollution loads in Lagos, Nigeria, outperforming other neural network models in accuracy and efficiency.

Contribution

It introduces the application of TLRN neural network model for estimating pollution loads, showing its superiority over other models in environmental pollution assessment.

Findings

TLRN outperformed other neural networks in MAE (0.14)

TLRN achieved a linear correlation coefficient of 0.84

ANNs are effective for environmental pollution estimation

Abstract

Industrial pollution is often considered to be one of the prime factors contributing to air, water and soil pollution. Sectoral pollution loads (ton/yr) into different media (i.e. air, water and land) in Lagos were estimated using Industrial Pollution Projected System (IPPS). These were further studied using Artificial neural Networks (ANNs), a data mining technique that has the ability of detecting and describing patterns in large data sets with variables that are non- linearly related. Time Lagged Recurrent Network (TLRN) appeared as the best Neural Network model among all the neural networks considered which includes Multilayer Perceptron (MLP) Network, Generalized Feed Forward Neural Network (GFNN), Radial Basis Function (RBF) Network and Recurrent Network (RN). TLRN modelled the data-sets better than the others in terms of the mean average error (MAE) (0.14), time (39 s) and linear correlation coefficient (0.84). The results showed that Artificial Neural Networks (ANNs) technique (i.e., Time Lagged Recurrent Network) is also applicable and effective in environmental assessment study. Keywords: Artificial Neural Networks (ANNs), Data Mining Techniques, Industrial Pollution Projection System (IPPS), Pollution load, Pollution Intensity.