Optimization of neural network classifiers by leveraging the sequential feature engineering for robust water quality prediction system

Rapid population growth increases water demand, intensifying extraction from wells and rivers. The Water Quality Index (WQI) assesses water suitability for drinking based on multiple parameters. Accurate assessment of pollution in water is imperative for effective management of water quality. The present research on the Neural Network-based Robust Water Quality Prediction System (NN-RWQPS) exploits the capabilities of neural networks and advances in feature engineering, positioning it at the forefront of WQI. Venturing into the new world of predictive modelling armed with four different neural network classifiers: Wide, Bilayer, Trilayer, and an Optimized Neural Network. Further the study harness the power of feature selection, deploying four distinct methods. A champion feature selection method is scientifically validated for each neural network, and then the neural networks are fine-tuned by training them across a range of feature dimensions, unveiling an empirically supported set of optimal features. Study advances water quality prediction using neural networks and feature engineering.

Proceedings on Engineering Sciences

Proceedings on Engineering Sciences

7

2683-4111

N/A

1

10