Performances of Different Machine Learning Algorithms for Predicting Saltwater Intrusion in the Vietnamese Mekong Delta Using Limited Input Data: A Study from Ham Luong River

Accurate forecasting of salinity intrusion has a vital role in water resource management to mitigate and prevent its adverse effects. However, monitoring of salinity presents great challenges because the task requires a number of information, such as hydrological, geomorphological data. The objective of this paper is to compare the performances of different machine learning algorithms for saltwater intrusion prediction using a limited number of input data. To achieve this goal, we tested the performances of five algorithms (i.e., Simple Linear, K-Nearest Neighbors, Random Forest, Support Vector Machine and a deep learning algorithm Long Short Term Memory) for predicting saltwater intrusion in the Ham Luong River in the Vietnamese Mekong Delta using only salinity monitoring data. We used Nash−Sutcliffe efficiency coefficient, Mean Absolute Error, and Root Mean Square Error to evaluate the performances of the above mentioned five models. Our results showed that Long Short Term Memory was the most accurate and efficient model, which implies that deep learning algorithms might be more efficient than machine learning algorithms in case of limited input data. Besides, the study also showed that performance of the linear model was insignificant compared to the non-linear algorithms. The results also revealed that saltwater intrusion forecasts could be achieved even in the limited data context. The present study provided a precise and simple tool for early warning of saltwater intrusion in the Vietnamese Mekong Delta.