Institutional-Repository, University of Moratuwa
Welcome to the University of Moratuwa Digital Repository, which houses postgraduate theses and dissertations, research articles presented at conferences by faculties and departments, university-published journal articles and research publications authored by academic staff. This online repository stores, preserves and distributes the University's scholarly work. This service allows University members to share their research with a larger audience.
 Research Publications |  Thesis & Dissertation |  E- Books |
 UoM Journal Publications |  UoM Conference Proceedings |  Articles published in Scimago's Q1 journals |  UoM Research Reports |  Other Articles authored by UoM staff |
Recent Submissions
item: Conference-Extended-Abstract
Model shifting unlearning: a scalable approach to data removal
(Department of Computer Science & Engineering, 2025) Pallewela, LCK; Athuraliya, CD
In the modern world digital age data has become the main driver of progress for several areas of life. However, the growing dependence upon data increased the alarm over the safety of people’s private rights with the advent of regulatory systems such as the European Union’s General Data Protection Regulation, which places importance upon the ”Right to be Forgotten” [1]. The law grants individuals the right to require the removal of personal details from company databases and presents serious challenges for machine learning (ML) algorithms that are drawing conclusions from huge data sources [2]. Conventional compliance solutions require extensive retraining of machine learning models for the removal of specific data, something that requires much resources. The solution is generally impossible for big models, highlighting the critical need for more efficient solutions. Current practices, such as the use of influence functions and the introduction of noise, attempt to solve the challenge; however, they often sacrifice model performance or are confronted with scalability [3]. Motivated by these challenges our study introduces Model Shifting Unlearning, a novel technique designed to efficiently remove specific data influences from ML models without necessitating full retraining. This method aims to identify and suppress neurons significantly impacted by the data to be forgotten, thereby maintaining overall model integrity and performance. The primary objective of this study is the development of a scalable unlearning framework that preserves model performance and testing of the efficiency of the framework compared with state of the art techniques [4]. With the solution of the great challenge of knowledge removal within machine learning, the study helps advance more ethically accountable and lawfully compliant AI systems.
item: Conference-Extended-Abstract
Storm track and intensity forecasting using a hybrid machine learning approach
(Department of Computer Science & Engineering, 2025) Lakshika, T; Ambegoda, T; Ekanayake, I; Meddage, P; Athuraliya, CD
Storms, including cyclones, hurricanes, and typhoons, pose significant threats due to their destructive winds and heavy rainfall, causing extensive damage to infrastructure, ecosystems, and human lives. Accurate prediction of storm paths and intensities is crucial for disaster preparedness and minimizing losses. Despite advances in meteorological science and computational modeling, accurately forecasting sudden changes in storm behavior, such as rapid intensification or unexpected weakening, remains challenging. Traditional numerical weather prediction (NWP) models integrate physical laws governing atmospheric conditions but struggle with uncertainties stemming from rapid weather variations, incomplete observations, and computational constraints. This study proposes a hybrid approach integrating sequence-based deep learning (Long Short-Term Memory (LSTM) and Transformers) with gradient boosting techniques (XGBoost, LightGBM) for predicting storm tracks and intensities. Additionally, physics-informed constraints such as storm kinetic energy, Coriolis force, and land interaction effects are incorporated to enhance the physical consistency and interpretability of the models. By combining machine learning models with fundamental meteorological principles, this work aims to improve the reliability and accuracy of cyclone forecasts. The main contributions of this study include: (1) the development of a multi-modal hybrid model combining deep learning and tree-based machine learning methods; (2) incorporation of physics-informed constraints to enhance model generalization and physical validity; and (3) rigorous evaluation of model performance using real-world cyclone datasets with robust metrics, including the Haversine distance for path prediction and Mean Absolute Error (MAE) for intensity estimation.
item: Conference-Extended-Abstract
The Impact of AI technologies on salary and industry demand for software engineering roles in Sri Lanka
(Department of Computer Science & Engineering, 2025) De Silva, P; Hewapathirana, I; Athuraliya, CD
The rapid integration of Artificial Intelligence (AI) technologies has transformed the global software engineering industry, and this change is no exception in Sri Lanka. This research study focuses on the impact of AI on salary trends, job role demands, and evolving skill requirements within Sri Lanka's software engineering sector. This study examined survey responses and analyzed LinkedIn job postings to identify the overarching trends: the surging demand for AI-centric roles, such as machine learning engineers, data scientists, and AI specialists, while the traditional roles like software developers are being redefined and specialized. It also investigates how dependence on AI is growing and focuses on the automatic code generation platforms, intelligent debugging assistants, and AI-driven testing frameworks, providing additional resources to accelerate software development process of software, enhancing productivity, and bringing quality to code. It calls for a complete set of new skills, which are machine learning, natural language processing, and AIbased software testing, in addition to bringing more efficiency into the workflow. Upskilling and educational reforms are crucial to addressing skill gaps created by AI in the software engineering landscape. It aims to provide actionable insights to industry stakeholders, educators, and policymakers on adapting to the AI-driven transformations necessary to maintain competitiveness of Sri Lanka in the software engineering workforce.
item: Conference-Extended-Abstract
A Systematic review on evaluating bias and equity in large language model (LLM) applications for patient communication in healthcare
(Department of Computer Science & Engineering, 2025) Kavishan, G; Arambepola, N; Athuraliya, CD
The rapid advancement of AI has transformed healthcare, with Large Language Models (LLMs) like ChatGPT and Med-PaLM enhancing patient communication by automating responses, summarizing medical documents, and aiding clinical decision-making, particularly in resource-limited settings. However, biases and inequities remain as critical challenges, undermining their effectiveness and fairness. Systemic biases can be culturally inappropriate, exacerbate healthcare disparities, and marginalize certain groups. Additionally, structural and economic factors contribute to these issues, necessitating urgent attention to demographic, cultural, and linguistic discrimination. This research aims to systematically evaluate these biases, their consequences, and potential solutions, offering recommendations to enhance AI driven healthcare communication systems, making them more just, efficient, and reliable.
item: Conference-Extended-Abstract
Internal risk rating model for finance institutes based on customer payment behaviour
(Department of Computer Science & Engineering, 2025) Jayasinghe, T; Watawana, T; Athuraliya, CD
With the implementation of International Financial Reporting Standards (IFRS) 9, Licensed Finance Companies (LFCs) are required to adopt a risk-averse approach to conservatively predict customer risk. As per the Central Bank of Sri Lanka’s (CBSL) directives on credit risk management for LFCs [1], credit facilities must be classified as either performing loans (PLs) or non-performing loans (NPLs) based on two criteria: (1) days past due (DPD) and (2) potential risk. Currently, LFCs predominantly rely on the DPD-based method. While this approach is simple to compute and interpret, it overlooks a customer’s payment history and behavioral trends. The second approach, classification based on potential risk—also known as Internal Risk Rating (IRR)—is not widely practiced. Organizations that have attempted this method primarily rely on demographic and geographical attributes. A significant drawback of this approach is that the data remains static, as it is recorded only at the time the credit facility is granted and does not capture subsequent changes. Additionally, LFCs collect vast volumes of business transaction data daily, including detailed payment histories. However, this rich dataset is primarily utilized for business monitoring through reports and dashboards rather than for predictive modeling. This study aims to develop an IRR model leveraging customer payment history data to improve risk
assessment.