Evaluation of Machine Learning Models in the Prediction of Water Quality Index for Selected Water Sources in Uyo, Akwa Ibom State, Nigeria

Why It Matters

Accurate, automated GWQI predictions enable faster, cheaper water‑resource management, crucial for regions facing monitoring constraints and public‑health risks.

Key Takeaways

• LassoLarsCV achieved R² = 1.0, zero error.
• Ridge Regression closely followed with R² = 0.9999.
• 48.3% of samples classified as poor water quality.
• Study used 13 physicochemical parameters from groundwater sources.
• ML models cut prediction time versus manual WAWQI calculations.

Pulse Analysis

Traditional water‑quality monitoring relies on labor‑intensive sampling and manual index calculations, which can delay decision‑making and inflate costs, especially in developing regions. Recent advances in artificial intelligence have opened pathways to automate these assessments, allowing agencies to process large datasets quickly and allocate resources more efficiently. By leveraging machine‑learning techniques, municipalities can shift from periodic testing to near‑real‑time surveillance, improving compliance with environmental standards and protecting public health.

The Uyo study collected groundwater samples from four clans and measured pH, temperature, EC, TDS, DO, BOD, alkalinity, acidity, hardness, chlorides, sulphate, phosphate and nitrate. Six regression models were trained on this multidimensional data, with LassoLarsCV delivering a flawless fit (R² = 1.0, RMSE = 0, MAE = 0). Its success highlights the power of regularized linear models to capture complex relationships while preventing overfitting, a common challenge in environmental datasets. Even the runner‑up, Ridge Regression, posted an R² of 0.9999, underscoring that well‑tuned linear approaches can rival more complex ensembles in this domain.

Beyond the academic proof‑of‑concept, the results signal a scalable solution for water‑resource managers worldwide. Integrating these predictive models with IoT‑enabled sensors could provide continuous GWQI updates, alerting officials to deteriorating conditions before they reach critical thresholds. Such proactive monitoring supports evidence‑based policy, optimizes treatment infrastructure investment, and ultimately safeguards communities that depend on safe groundwater supplies. As climate variability intensifies, AI‑driven water‑quality forecasting will become an essential component of resilient environmental stewardship.