Machine Learning and Deep Learning Models for Water Quality Prediction with SHAP-Based Explainable AI

The prediction of water quality has an important role to play in ensuring public health safety and management of the environment. In India, rising pressure on river systems makes water quality prediction necessary. Predictive models built on machine learning algorithms have high efficiency levels, but because of being black boxes, lack of confidence in adopting such models prevails. In this context, three predictive models—SVM, LR, and LSTM networks—have been compared based on the Indian River system dataset to predict water quality. SHapley Additive exPlanations (SHAP) is utilized to interpret predictions made by models and determine those physicochemical parameters that affect the outcome of each prediction. It has been found that LSTM is the most effective of the three models considered in predicting water quality, and it does so due to its ability to capture temporal dependencies in water quality parameters, while SHAP determines the reasons behind each prediction.