UoK Repository

Hosted by UoK@ICTC

Communities in DSpace

Select a community to browse its collections.

Now showing 1 - 5 of 15

Recent Submissions

EVALUATING THE EFFICACY OF THE SOFTWARE STARTUP ECOSYSTEM IN SRI LANKA: THE ROLE OF ACCELERATORS, INCUBATORS, INVESTORS, AND GOVERNMENT
(The Library, University of Kelaniya, Sri Lanka., 2025) Nagahawatte, V. G. D.; Wijayanayake, J.
The burgeoning software startup ecosystem in Sri Lanka holds significant potential to drive national economic growth and technological innovation. However, this ecosystem faces critical challenges that hinder its global competitiveness and scalability, including inadequate infrastructure, limited funding opportunities, and a scarcity of skilled talent. This research evaluates the efficacy of the Sri Lankan software startup ecosystem, emphasizing the pivotal role of incubators, accelerators, and government bodies. Employing a quantitative approach, the study identifies key success factors, challenges, and actionable strategies to enhance ecosystem performance. The research employs Partial Least Squares Structural Equation Modeling (PLS-SEM) to analyze data collected from stakeholders, including startup founders, employees, incubators, accelerators, and investors. The analysis revealed that advisory and mentorship, as well as corporate partnerships, had a statistically significant positive impact on the efficacy of the software startup ecosystem. In contrast, access to capital, innovation and R&D, and the regulatory environment did not show significant direct influence within the current Sri Lankan context. The study aims to contribute to the strategic development of Sri Lanka's software startup ecosystem, enabling it to achieve sustainable growth and align with global benchmarks.
DEVELOPING A PRIORITY MODEL FOR THE MATURITY OF AI-DRIVEN ERP INTEGRATION IN SRI LANKAN CONTEXT
(The Library, University of Kelaniya, Sri Lanka., 2025) Galwaththa, G. V. N. P.; Oruthotaarachchi, C. R.
The integration of Artificial Intelligence (AI) with Enterprise Resource Planning (ERP) systems is emerging as a critical driver of digital transformation, enabling organizations to enhance operational efficiency, optimize decision-making processes, and maintain a competitive edge. Despite its potential, the successful integration of AI into ERP systems is influenced by a variety of organizational, technological, and leadership factors, which can differ significantly across regions and industries. This study investigates the key factors affecting AI-ERP integration within the Sri Lankan business environment, aiming to develop a comprehensive priority model that reflects the maturity of AI integration in this specific context. A thorough literature review was conducted to identify relevant success factors, which were then validated through consultations with industry experts. These validated factors were incorporated into a structured questionnaire distributed to a targeted sample population within the Sri Lankan corporate sector. The collected data were analyzed using Partial Least Squares Structural Equation Modeling (PLS-SEM) to determine the significance and impact of each factor. Based on the analysis, a five-level priority model was developed to represent the maturity stages of AI-ERP integration. The model provides a structured framework that highlights the most critical areas organizations should focus on to ensure successful integration. This research offers valuable insights for Sri Lankan enterprises seeking to adopt AI-enabled ERP systems and provides a strategic benchmark for assessing integration readiness. Furthermore, the findings serve as a foundation for future academic research and contribute practical implications for decision-makers and technology leaders in developing effective AI-ERP integration strategies.
PREDICTIVE MACHINE LEARNING FRAMEWORK FOR POTENTIAL DELAY
(The Library, University of Kelaniya, Sri Lanka., 2025) Kariyawasam, J.; Rajapakse, C.; Dissanayake, M.
Delays relating to order deliveries in elastic manufacturing can significantly disrupt production timelines, affect supply chain efficiency, and diminish customer satisfaction. This research presents a machine learning-based framework to address these challenges by predicting manufacturing delays, explaining their underlying causes, and estimating delay durations to improve operational reliability. The study employs a three-layer architecture: a binary classification layer to predict whether an order delivery delay will occur, an explainability layer using Local Interpretable Model-Agnostic Explanations (LIME) to provide insights into each prediction, and a regression layer to estimate the duration of the delay and the expected delayed delivery date. The dataset, spanning more than three years (from 2021-01-08 to 2024-06-24), includes 37,411 order records and 75,723 delivery records, reflecting the complexities of elastic manufacturing order fulfillment. Extensive data preparation was conducted to standardize formats, handle missing values, and normalize features. Classification models such as Logistic Regression, XGBoost, and Neural Networks were utilized for classification tasks, while Linear Regression and XGBRegressor were employed for regression tasks. The findings demonstrate that the proposed framework not only accurately predicts delays but also offers actionable insights into their drivers, enabling proactive decision-making. By integrating predictive power with explainability, this study contributes to the advancement of intelligent supply chain management systems.
SRI LANKA’S TRAIN-ELEPHANT COLLISION CRISIS: A REVIEW OF POLICY AND TECHNOLOGICAL INTERVENTIONS
(The Library, University of Kelaniya, Sri Lanka., 2025) Kalhara, T. G. D.; Fernando, K.
Human-elephant conflict (HEC) is a long-standing issue in Sri Lanka, where expanding human settlements, agriculture, and infrastructure have increasingly encroached upon elephant habitats. This conflict has intensified in recent years, resulting in significant casualties on both sides, including damage to crops, property, and lives. However, increasing human expansion and infrastructure development have led to frequent human-elephant conflicts (HEC). Sri Lanka's railway infrastructure intersects with critical elephant habitats, creating a dangerous overlap that has resulted in increasing train-elephant collisions, particularly from 2020 to 2025. These incidents pose a severe threat to the country's biodiversity, especially its endangered elephant population, while also causing infrastructure damage and operational disruptions. This study explores the root causes, statistical trends, policy actions, and technological interventions aimed at reducing train-elephant collisions across high-risk railway corridors such as the Trincomalee, Batticaloa, and Northern lines. This study reviews the causes, statistical trends, government policies, and technological innovations aimed at addressing the issue. While policy interventions such as train speed limits, schedule adjustments, and vegetation clearance have been implemented, their effectiveness has been limited by enforcement challenges and infrastructural constraints. Meanwhile, promising technological solutions such as AI-powered detection systems, Arduino-based sensors, Fiber Bragg Grating (FBG) sensing technology, and high-frequency sound deterrents have shown potential in detecting elephant movement and issuing early warnings to prevent collisions. Moving forward, it is essential for policymakers, conservationists, and technologists to work together in designing sustainable solutions that protect both Sri Lanka's national heritage the elephant and the safety of its railway system. Only through integrated and innovative approaches can a balance be struck between development and wildlife conservation.
REAL VS. FAKE FACE CLASSIFICATION USING DEEP LEARNING-BASED CONVOLUTIONAL NEURAL NETWORKS
(The Library, University of Kelaniya, Sri Lanka., 2025) Shanuka, A.; Weerasinghe, M.; Adikari, T.; Sandasara, D.; Mahanama, T.
The widespread use of artificial intelligence (AI) has accelerated the creation of hyperrealistic synthetic images, commonly referred to as deepfakes. These AI-generated faces present growing challenges in domains such as security, media verification, and digital identity. This study presents a primary research effort that investigates the classification of real and AI-generated facial images using a deep learning-based convolutional neural network (CNN). Unlike previous works that solely rely on existing public datasets, this study integrates a curated set of AI-generated and real faces from Kaggle with a custom-collected image set to enhance dataset diversity and realism. The experimental CNN architecture was developed using the Keras Sequential API and consists of four convolutional layers with ReLU activation, four max-pooling layers, one batch normalization layer, dropout regularization, and two fully connected dense layers. The training process incorporated data augmentation techniques such as rescaling, shearing, zooming, and horizontal flipping to improve generalization and mitigate overfitting. The model was compiled using the Adam optimizer and binary cross-entropy loss, with accuracy used as the primary metric. The model was trained for 25 epochs, achieving a test accuracy of approximately 82% on the final evaluation. Performance metrics indicate high precision (83.08%) and recall (82.44%), as reflected in the classification report and confusion matrix. These results confirm the model's effectiveness in accurately differentiating between real and AI-generated faces. Model performance was monitored through training and validation curves, revealing consistent accuracy gains and manageable loss fluctuations. In addition to model training, feature map visualizations were generated to illustrate how the CNN processes facial structures across layers. The findings suggest that intermediate convolutional outputs successfully capture critical features like facial symmetry, edge clarity, and texture anomalies, which are typically absent or uniform in synthetic faces. The inclusion of a custom dataset and extended experimental controls, such as validation monitoring and architectural tuning, elevate this work from a secondary to a primary research contribution. This study highlights the practical application of CNNs in deepfake detection and proposes future extensions to real-time video classification and architectural comparisons involving ensemble or transfer learning models.