Classification of water quality using artificial neural network

• المؤلف الرئيسي: Sulaiman, Khadijah
• التنسيق: أطروحة
• اللغة: English; English; English
• منشور في: 2020
• الموضوعات: QA76 Computer software
• الوصول للمادة أونلاين: http://eprints.uthm.edu.my/1013/1/24%20KHADIJAH%20SULAIMAN.pdf; http://eprints.uthm.edu.my/1013/2/KHADIJAH%20SULAIMAN%20COPYRIGHT%20DECLARATION.pdf; http://eprints.uthm.edu.my/1013/3/KHADIJAH%20SULAIMAN%20WATERMARK.pdf; http://eprints.uthm.edu.my/1013/

Deterioration in water quality has triggered many countries to initiate serious mitigation efforts because it is essential to prevent and control water quality pollution and to implement regular monitoring programmes to preserve the environment. Hence, the water quality index (WQI) developed by Department of Environment (DOE) Malaysia has been used to classify water quality in Malaysia for the past few decades. However, another method has emerged for classifying water quality, for example soft computing approach. Therefore, this research aims to use the artificial neural network (ANN) algorithm to classify water quality at Pontian Kechil, Batu Pahat and Muar river. Concentrations of pH, suspended solids (SS), dissolved oxygen (DO), chemical oxygen demand (COD), biological oxygen demand (BOD), and ammoniacal-nitrogen (NH3-N) were measured in situ and via laboratory analysis. Concentrations of these six parameters were used in the mathematical equation of the DOE-WQI technique and as input variables in the ANN database system. Based on the average WQI values of 35 stations for each river, it is shown that Pontian Kecil and Batu Pahat river were categorised in class III with WQI values of 71.2 and 71.5 respectively. Meanwhile, Muar river were categorised in class II with WQI values of 76.8. The performance of ANN model was identify based on the classification accuracy, sensitivity, precision, and root mean square error (RMSE). Then, the model performance was compared with the k-NN and Decision Tree models. The results obtained after training and testing the network showed that ANN performed better than to k-NN and Decision Tree model with the values of accuracy, sensitivity, precision and RMSE is 95.24%, 93.51%, 94.43% and 0.207 respectively. The ANN produced highly accurate water quality classification. In addition, it is more flexible than existing approaches and can be implemented easily and quickly.