An Efficient Method for Water Quality Prediction for Ungauged River Catchment Under Dual Scenarios Based on CNN-BiRNN-A Approach

All life on Earth depends on water, which makes up more than 80% of its surface. Harmful water pollution has emerged as a result of the fast deterioration of water quality caused by increasing industrialization and urbanization. Most water quality estimations need time-consuming and costly statistical calculations, rendering current real-time monitoring efforts useless. There is a predetermined sequence in which to carry out preprocessing, feature extraction, and model training. An important part of the preprocessing phase of data analysis is data normalization using the z-score approach. This step is vital for getting high-quality findings. The suggested method improved the accuracy of water pollution process classification by employing VGG for feature extraction. The model was trained using a CNN-BiRNN-A technique. In comparison to state-of-the-art CNN and BiRNN, our approach seems fresh. With a 97.63 % accuracy percentage, the numbers demonstrated a notable enhancement.