Explaining the Importance of Water Quality Parameters for Prediction of the Quality of Water Using SHAP Value

Abstract

Water is key to life on planet Earth, and hence, maintaining water quality is a critical issue in contemporary times. The water quality index decides the quality of drinking water. The presented work first explores different machine learning algorithms on the already collected water samples to decide the water quality and then applies the coalition game theory-based SHapley Additive exPlanations (SHAP) approach to decide the significance of each parameter in deciding the class of water sample based on quality. The potential of popular algorithms like K-NN, support vector machine, decision tree, etc. are being explored to find out the quality of water samples. All the machine learning algorithms used in the work give over 80% accuracy while the performance of neural network is 96% proving to be the best among all other algorithms. The presented work demonstrates the model agnostic, coalition game theoretic SHAP value-based method for explaining the importance and impact of each of the given parameter pH, HCO3-, Cl-, NO3-, F-, Ca, Mg, Na, Ec, etc. in deciding the quality of the water.