Salinity intrusion is considered to be one of the important environmental degradations, causing negative effects on agricultural development, which is increasingly serious due to climate change and sea level rise. The Red River Delta, considered the rice granary of Vietnam, is among the regions most vulnerable to salinity intrusion. Although this region is important for the country's food security, however, less study has been conducted to predict salinity intrusion. The objective of this study is to evaluate the salinity intrusion based on hybrid machine-learning alogrithms, namely Xgboost (XGB), random forest (RF), LightGBM, Xgboost-Decision tree (XGB-DT), random forest-Decision tree (RF-DT), LightGBM-Decision tree (LightGBM-DT), Xgboost-Linear regression (XGB-LR), random forest-linear regression (RF-LR), LightGBM-Linear regression (LightGBM-LR) in the Red River Delta in Vietnam. Hourly water level, precipitation, and temperature dataset from 2014 to 2023 were used to predict salinity intrusion. The results showed that all proposed models were effective in predicting salinity intrusion in a Red River Delta, with the value of R² > 0.8. Among them, the Xgboost-DT model was more accurate than other models with an R2 score of 0.86. The salinity at the Ba Lat station reached its highest peak of 23-24 g/l, and all proposed models captured the trend of salinity in the study area, in which the LightGBM model and its hybrid model provided the highest precision in terms of time and intensity of salinity. The outcomes of this study play an important role in supporting decision-makers or farmers in establishing effective measurements and optimizing water resource management to reduce the effects of salinity intrusion on agriculture.