Open AccessPREDICTION OF THE LEVEL OF WATER QUALITY INDEX USING ARTIFICIAL NEURAL NETWORK TECHNIQUES IN MELAKA RIVER BASIN
Author(s) -
Ang Kean Hua
Publication year - 2020
Publication title -
malaysian applied biology/malaysian applied biology journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.153
H-Index - 8
eISSN - 2462-151X
pISSN - 0126-8643
DOI - 10.55230/mabjournal.v49i1.1656
Subject(s) - water quality , ammoniacal nitrogen , artificial neural network , biochemical oxygen demand , chemical oxygen demand , linear regression , correlation coefficient , statistics , data set , index (typography) , environmental science , coefficient of determination , sampling (signal processing) , test set , variables , mathematics , environmental engineering , computer science , machine learning , ecology , filter (signal processing) , wastewater , world wide web , computer vision , biology
Artificial Neural Network (ANN) techniques were used to develop and validate water quality by predicting the Water Quality Index (WQI) in Melaka River Basin, Malaysia. Nine sampling stations were monitored in total. ANN techniques were applied for testing and developing the water quality prediction based on two sets of data. In the first data set, the independent water quality of six variables was used as input into ANN for trained, test and validated samples. In the second data set, a combination between Multiple Linear Regression (MLR) and ANN indicating only Chemical Oxygen Demand (COD), Biochemical Oxygen Demand (BOD), Suspended Solid (SS), and Ammoniacal-Nitrogen (AN) are accounted for training, testing and validating in modeling the water quality. Generally, MLR is used to exclude the lowest value invariance of independent variables, while rejecting the Dissolved Oxygen (DO) and pH. Based on the result of the correlation coefficient, the second set data (0.89) is marginally better than the first set data (0.87). These circumstances stated that predictions for WQI using ANN are acceptable, and the result is better when the variables of DO and pH are eliminated.