Deep Ensemble-based Water Quality Index Prediction and Classification using Randomized Low-Rank Approximation

The drinking water and the planet’s well-being depend on its purity. Many different industries and sectors rely on reliable access to clean water for their operations. Indicators like the water quality index (WQI) are essential for efficient water management. In recent years, several qualities of water has been compromised by pollutants. Because of this, limiting water pollution has elevated the significance of modeling and predicting water quality. In this study, state-of-the-art AI algorithms are created to provide predictions regarding WQI and water quality classification (WQC). Predicting groundwater quality is useful for reducing water pollution. This research provides a novel technique with novel groundwater quality index forecasting factors based on double-edged Bi-LSTM that take into consideration external factors with a significant impact on the water quality indices. Groundwater abstraction was conducted, and the effective factors of the groundwater quality index were found to be rainfall rate, temperature, and humidity. The "f-score" test is used to determine the optimal neural network architecture. When compared to the actual value, most predictions of the indicator of water quality accurate and reasonable. Four statistical metrics, including Accuracy, Recall, F1-score were used to evaluate the accuracy of the AI predictions and the GWQI calculations. According to the results of the study, ensemble deep learning (DL) algorithms may be utilized to accurately estimate and forecast the Water Quality Index.