Research Article
Explainable AI-Based Water Quality Prediction System
@INPROCEEDINGS{10.4108/eai.28-4-2025.2357836, author={Jayarama Krishna Challagundla and Alla Nithin Reddy and Arul Elango and Surya Vipparla and Manoj Dasari and Rodda Maheswara Reddy}, title={Explainable AI-Based Water Quality Prediction System}, proceedings={Proceedings of the 4th International Conference on Information Technology, Civil Innovation, Science, and Management, ICITSM 2025, 28-29 April 2025, Tiruchengode, Tamil Nadu, India, Part I}, publisher={EAI}, proceedings_a={ICITSM PART I}, year={2025}, month={10}, keywords={water quality explainable ai (xai) machine learning shap lime potability prediction dimensionality reduction linear discriminant analysis (lda) classification algorithms}, doi={10.4108/eai.28-4-2025.2357836} }- Jayarama Krishna Challagundla
Alla Nithin Reddy
Arul Elango
Surya Vipparla
Manoj Dasari
Rodda Maheswara Reddy
Year: 2025
Explainable AI-Based Water Quality Prediction System
ICITSM PART I
EAI
DOI: 10.4108/eai.28-4-2025.2357836
Abstract
Access to safe and clean drinking water is a global issue, especially in developing countries. Traditional water testing methods are always slow, costly and not available in remote areas. To overcome this limitation, we proposed Explainable AI based Water Quality Prediction System which uses machine learning and explainable models to classify whether the water is drinkable or not. We used a dataset which contains 3000 water samples, each sample contained six core parameters: pH, Turbidity (NTU), Chlorides (mg/L), Dissolved Solids (mg/L), Alkalinity (as CaCO3), and Fluorides (mg/L). While preprocessing we converted the categorical values into numerical values and applied normalization using StandardScaler. Several classification algorithms were developed and evaluated, including Random Forest, Logistic Regression, K-Nearest Neighbors (KNN), Decision Tree, Gaussian Naïve Bayes, and Support Vector Machine (SVM). Furthermore, dimensionality reduction was performed using Linear Discriminant Analysis (LDA), with performance comparisons made before and after reduction. For model interpretability and trust, explainable AI techniques such as LIME and SHAP were utilized to highlight feature importance as well as provide insight into the decision process. The proposed system is highly accurate and interpretable and is therefore suitable for implementation in real-world, resource-constrained environments where accurate and interpretable water quality prediction is key to public health.
