Water quality monitoring is essential for sustainable resource management. This study applies Machine Learning (ML) techniques to predict Total Dissolved Solids (TDS) in the Tigris River at three different locations in Iraq. Using Artificial Neural Networks (ANN), Random Forest (RF), and Linear Regression (LR), predictive models were developed based on key water parameters, including conductivity, turbidity, alkalinity, hardness, pH, and major ions. Feature selection methods such as Akaike Information Criterion (AIC) for LR, permutation importance for RF, and the Olden method for ANN identified conductivity as the strongest predictor of TDS. Among models, RF showed the lowest prediction error outperforming ANN and LR. The study showed AI-driven predictive modelling as a powerful tool for real-time water quality assessment. With increasing TDS levels in Iraq due to pollution and climate effects, ML models provide a reliable framework for early contamination detection and sustainable water management.