https://www.um.edu.mt/library/oar/handle/123456789/135418 2026-05-04T06:33:53Z https://www.um.edu.mt/library/oar/handle/123456789/146015 Title: Investment portfolio through evolutionary algorithm Abstract: This dissertation investigates the application of evolutionary algorithms—specifically the Genetic Algorithm (GA) and Simulated Annealing (SA)—for portfolio optimisation within the S&P 500, addressing the limitations of traditional models. The methodology uses Gower’s distance to handle mixed numerical and categorical data, allowing for the construction of factor-aligned portfolios based on Growth, Value, and Quality dimensions. The primary optimisation objective is to maximise the Sortino Ratio, focusing on downside-risk-adjusted returns. The methodology employs composite feature engineering to rank stocks across Growth, Value, and Quality dimensions, followed by distance-based clustering and anomaly detection to reveal market structures. GA and AGA are then applied with objective functions designed to maximise the Sortino Ratio, which emphasises downside-risk-adjusted returns. Hyperparameters such as population size, mutation rate, crossover probability, and annealing temperature schedules are tuned to balance exploration and exploitation. Empirical evaluation demonstrates that both GA and AGA generate highly diversified portfolios with competitive performance. The GA-optimised maximum Sortino portfolio achieved an annualised return of 18.95%, a Sortino Ratio of 1.0999, and a beta of 0.98, indicating strong returns with reduced downside risk relative to the market. Comparative analysis reveals that AGA converges faster and achieves marginally superior downside protection, validating its advantage in complex search landscapes. Complementary tools such as similarity maps, hierarchical clustering, and a diversification recommender further enhance interpretability and practical applicability. The results underscore the potential of evolutionary algorithms to construct robust, risk-aware investment portfolios that go beyond linear optimisation frameworks. By combining factor-based insights with evolutionary optimisation, this work contributes to the growing literature on computational finance and demonstrates actionable applications for institutional and retail portfolio managers. Description: M.Sc.(Melit.) 2025-01-01T00:00:00Z https://www.um.edu.mt/library/oar/handle/123456789/146014 Title: Hand-gesture recognition based on sEMG using deep learning architectures Abstract: Hand gesture recognition is a key area in non-verbal communication, enabling intuitive and touchless interaction between individuals and digital systems. As non-verbal communication plays a vital role in human interaction, hand gesture recognition systems can improve accessibility and increase communication efficiency. Over time, hand-gesture recognition has been considered more important, and this can be achieved by using surface electromyography (sEMG) signals. SEMG is a type of electromyography (EMG) procedure where the signals are recorded on the skin surface rather than within the muscle. The main problem with sEMG signals is that there are several physiological processes in the skeletal muscles underlying their generation. This is the main reason gesture recognition using an sEMG is a non-trivial task. Noise is also a contributing factor to the problem with sEMG signals. A dataset is created with 30 participants and 10 communication hand gestures that is then split between training, validation, and testing. To create the dataset, sEMG signals are collected via a controlled experiment using a hand gesture recording device such as the Myo armband. This study explores deep learning algorithms for hand gesture classification and evaluation. The implementation of real-time hand gesture recognition is studied using Convolutional Neural Networks (CNNs) and Long Short Term Memory (LSTM). The research examines how these models can adapt to dynamic movement and positioning, which may affect recognition accuracy. A key objective of this study is to provide a reliable and efficient manner in predicting hand gestures, making it applicable in various fields. One of those fields is verbal communication throughout the day. Experimental results demonstrate the reliability of integrating gesture recognition into an information system that predicts hand gestures in real time, offering improved accessibility and communication support. By optimising feature selection and model performance, this research contributes valuable insights for advancing gesture-based predictive systems, by achieving a net result of 80.67% for the recognition of 10 hand gestures. This dissertation enhances the field of non-verbal communication through gesture recognition, paving the way for more sophisticated and accessible interaction technologies. Description: M.Sc.(Melit.) 2025-01-01T00:00:00Z https://www.um.edu.mt/library/oar/handle/123456789/145526 Title: Unified load balancing strategies for enhanced cloud computing solutions Abstract: Cloud computing offers scalable, on-demand resources that enable a variety of services and applications. Effective load balancing in cloud environments is essential for maintaining performance and Quality of Service (QoS). These environments present complex, dynamic conditions that make efficient load balancing challenging. Many existing algorithms focus on single-objective optimisation, such as minimising response time, which often results in trade-offs and inefficiencies when dealing with unpredictable workloads. This dissertation tackles these inefficiencies by introducing a unified, multi-objective load balancing strategy that combines Ant Colony Optimisation (ACO) and Genetic Algorithm (GA) techniques. The hybrid ACO-GA algorithm is implemented within the CloudAnalyst simulation environment, leveraging ACO’s rapid local search and GA’s global exploration capabilities to dynamically balance workloads across cloud resources. Extensive simulation experiments demonstrate that the proposed hybrid approach significantly improves key QoS metrics compared to both conventional and state-of-the-art load balancers. The ACO-GA consistently achieved substantially lower average response times and improved load distribution relative to traditional algorithms. For example, under light workloads it reduced mean response time by roughly 50% versus Round Robin and 40% under heavy loads. The hybrid method also outperformed modern heuristics, sustaining about 8–10% faster response than advanced metaheuristic policies while shortening data centre processing delays. These gains were accompanied by more efficient resource utilisation, as the algorithm prevented server overloading and underutilisation through balanced task allocation. Notably, performance improvements persisted across both low and high demand scenarios, highlighting the algorithm’s robust adaptability to dynamic cloud conditions. Overall, the results affirm that this unified ACO-GA strategy effectively addresses the limitations of single-objective approaches, offering a significant enhancement in cloud service performance, resource utilisation and QoS. Description: M.Sc. ICT(Melit.) 2025-01-01T00:00:00Z https://www.um.edu.mt/library/oar/handle/123456789/144338 Title: Towards the efficient adaptation of offline physically based methods for real-time rendering Abstract: Rendering physically accurate caustics in real-time remains a persistent challenge due to their complex light interactions and high-frequency features. This thesis presents a comprehensive exploration into adapting oìine physically based rendering techniques for real-time caustic synthesis on modern GPU architectures. Central to this work is CandelaDXR, a GPU-accelerated light tracer that employs novel importance sampling strategies to improve convergence speed and visual adelity for caustics. By generating dynamic probability distribution functions conditioned on scene geometry, camera parameters, and material properties, CandelaDXR prioritises specular interactions and focuses sampling eêorts on perceptually relevant regions. To support this system, two auxiliary tools were developed: Anvil, a modular visual debugging platform for rendering pipelines, and Forge, an evaluation framework designed to facilitate reproducible, cross-system comparisons. These tools provide both insight and rigour in diagnosing rendering artefacts and validating algorithmic improvements. Additionally, the thesis introduces a spectral denoising pipeline tailored to the distinct characteristics of caustic signals, demonstrating the eêectiveness of Fourier, wavelet and curvelet-based transforms in preserving detail while reducing noise. Quantitative results across multiple scenes and viewpoints reveal signiacant performance gains, noise reduction, and perceptual improvements in CandelaDXR over baseline methods. Collectively, this work contributes a uniaed architecture for real-time caustic rendering, debugging, and evaluation, offering practical advances in both rendering theory and implementation for real-time physically based graphics. Description: Ph.D.(Melit.) 2025-01-01T00:00:00Z