Prediction Water Potability
Mar 2024 - May 2024 (3 months)
Uses machine learning to classify water samples as potable or non-potable based on chemical and physical properties such as pH, turbidity, hardness, chloramines, sulfate, and conductivity. Preprocessed a dataset by handling missing values using class-based mean imputation and addressed class imbalance with SMOTE. Trained and evaluated three models—Random Forest, XGBoost, and LightGBM—and found that Random Forest performed best with 71.38% accuracy. The project demonstrates how AI can support water quality assessment and public health decision-making, especially in developing regions where water contamination is a major concern.