River Water Quality Prediction in Malaysia Based on Extra Tree Regression Model Coupled with Linear Discriminant Analysis (LDA)

Djarum, DH; Ahmad, Z; Zhang, J

doi:10.1016/B978-0-323-88506-5.50230-8

River Water Quality Prediction in Malaysia Based on Extra Tree Regression Model Coupled with Linear Discriminant Analysis (LDA)

Lookup NU author(s): Dr Zainal Ahmad, Dr Jie Zhang ORCiD

Downloads

Full text for this publication is not currently held within this repository. Alternative links are provided below where available.

Abstract

© 2021 Elsevier B.V.Poor river water quality as a result of rapid urbanization has led to various disease outbreak and destruction of wildlife ecosystem. Although continuous water quality monitoring has been proposed as the immediate solution, the complexity and expensive nature of such system often offset its purpose. Therefore, numerous studies have been made to forecast water quality utilizing different machine learning algorithms. In this paper, a comparative study was carried out to develop an efficient water quality index (WQI) prediction model based on more attainable monitoring parameters. Three different variation of ensemble decision tree models were analysed and compared, namely: Random forest regression (RF), extra tree regression (ETR), and decision tree + AdaBoost (BTR). These models were coupled with principle component analysis (PCA) and linear discriminant analysis (LDA) to reduce the dimensions of the input data. The results show that (ETR-LDA) model outperform the other ensemble tree models with an R2 value up to 0.88. The ETR-LDA combination consistently score a higher R2 values even when trained with a much smaller input dimension (i.e. 3) resulting in a faster training time. This model could positively contribute towards the long-term water quality management effort in a cost-effective manner.