Improving Water Quality Index prediction for water resources management plans in Malaysia: application of machine learning techniques

Proper modeling of groundwater quality is an important planning and decision-making tool in water resources management and environment. Due to the fact that forecasting and determining groundwater quality helps in planning and managing groundwater, in this study we tried to use intelligent models with different functions to predict Water Quality Index (WQI). The traditional methods to compute the WQI is time consuming and have errors in sub-indexes derivations. Three different Machine learning (ML) techniques are applied to a 1080 data of Kelang River which collected from 27 sampling sites in the dry season (January/February, 2014) and wet season (October/November, 2014) to predict WQI. For modeling with ML algorithms six effective input parameters are considered and 70% of all data were used for training stage and 30% for testing stage. All three models demonstrated a good performance in prediction of WQI but the LSTM algorithm with R2 = 0.986, RMSE = 1.383, MAE= 0.924 and NSE = 0.992 showed the best performance than those of other models.