Mathematical modeling, Optimization,  infection diseases, machine learning, deep learning, physics-informed neural network, graph neural network, dynamical systems, operational research and Software engineering.

Data-Driven Algorithms for Forecasting Critical Care Demand during the COVID-19 Pandemic:

This project brings together computational modelling, epidemiology, and machine learning to improve how healthcare systems anticipate and manage demand during major infectious disease outbreaks. The work focuses on the challenge of forecasting COVID-19 hospitalisation and ICU demand and using these forecasts to support more effective planning and allocation of limited medical resources. Pandemic surveillance data are often noisy, incomplete, and uneven across regions, which makes the forecasting problem both complex and computationally demanding.

The thesis answers this challenge by developing a suite of advanced data-driven and hybrid modelling approaches grounded in real NHS England data. It first evaluates a range of deep learning architectures, such as RNNs, LSTMs, and Seq2Seq models, to understand how temporal irregularities, feature design, and hyperparameter choices influence national-level forecasting performance. Building on this, the research introduces a hybrid framework that combines Physics-Informed Neural Networks with a mechanistic SEIaIsHCRD model to recover hidden epidemic states and time-varying parameters in settings where observations are incomplete. Finally, the thesis proposes MSAGAT-Net, a multi-scale graph attention network designed to capture both spatial and temporal structure in regional transmission patterns, which leads to substantial improvements in multi-horizon forecasting accuracy.

Together, these contributions show how complex and imperfect epidemic data can be transformed into reliable intelligence for decision-makers. By linking modern machine learning with epidemiological theory, the thesis provides practical modelling tools that support stronger preparedness, better resource planning, and more resilient health system responses in future public health emergencies.

Plant disease detection: A cross-platform mobile application using Artificial Neural Networks