Enhancing Water Potability Prediction using Sparse Attention-Based Deep Learning Models

Pakeer, Saikiran Reddy (2025) Enhancing Water Potability Prediction using Sparse Attention-Based Deep Learning Models. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (777kB) | Preview
Preview	PDF (Configuration Manual) Download (557kB) | Preview

Abstract

Reliable prediction of water potability is essential for safeguarding public health and supporting sustainable resource management. However, real-world water quality data is often subject to noise, sensor inaccuracies, and missing values, which can degrade model performance. This study proposes the use of a sparse attention-based deep learning model to enhance the robustness and generalizability of water quality prediction systems. By introducing controlled Gaussian noise at varying intensities to the training data, the model’s capacity to maintain predictive accuracy under noisy conditions is systematically evaluated. The proposed approach leverages the TabNet architecture, which employs feature-wise attention to focus selectively on the most informative inputs. This mechanism enables effective learning even when data quality is degraded. Experimental results demonstrate that the sparse attention framework remains stable across multiple perturbation levels, underscoring its potential for deployment in real-time and uncertain environments. The findings suggest that incorporating sparse attention into predictive systems can significantly strengthen their resilience, making them more suitable for real-world water quality monitoring applications.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Horn, Christian UNSPECIFIED
Subjects:	R Medicine > RA Public aspects of medicine > RA0421 Public health. Hygiene. Preventive Medicine G Geography. Anthropology. Recreation > GE Environmental Sciences > Environment Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	02 Jul 2026 14:16
Last Modified:	02 Jul 2026 14:16
URI:	https://norma.ncirl.ie/id/eprint/9443

Actions (login required)

View Item