Physical Sciences and Mathematics Commons^™

Premier League Results Predictions, Laura Consuegra

ICT

This project applies machine learning techniques to predict outcomes in the English Premier League, one of the most prestigious and widely followed football competitions worldwide. By analysing historical and real-time match data, including performance metrics such as goals scored, shots on target, and cards received, predictive models are developed to forecast match results with higher accuracy. The study evaluates the effectiveness of these models and explores the influence of key statistical features on team performance. The findings aim to provide strategic insights for fans, bookmakers, and coaching staff, supporting performance evaluation, tactical decision-making, and a deeper understanding of the factors …

A Comparative Analysis Of The Performance Of Customised And Pre-Trained Convolutional Neural Networks In The Classification Of Facial Expressions., Guilherme Soares Da Costa

ICT

Facial expression recognition (FER) is a highly relevant problem in the field of computer vision, with applications in various areas. Although this topic has made significant progress in recent years with deep learning, facial expression recognition continues to face significant challenges when faced with a noisy, imbalanced database, and low-resolution images, as is the case with the widely used FER2013 database. This work investigates the usability of different approaches based on Convolutional Neural Networks (CNNs) for the task of facial expression classification.

First, a custom CNN was developed and trained from scratch, using FER2013 as database. Subsequently, three well-known pre-trained …

Clustering And Predictive Modelling Of Cryptocurrencies: An Empirical Study Of Lead–Lag Dynamics And Forecasting Performance., Viktor Varga

ICT

Cryptocurrency markets pose analytical challenges due to their volatility, non-stationarity, and complex interdependencies. Traditional econometric models have struggled with these features, while machine learning models often face overfitting and limited interpretability. This thesis has developed an integrated framework combining clustering, lead–lag detection, and predictive modelling to examine whether structural and temporal dependencies in cryptocurrency markets can improve forecasting performance.

Hourly OHLCV data for a selection of liquid cryptocurrencies from the Binance API over a thirteen-month period has been analysed. Dynamic Time Warping (DTW) has been used to identify asset clusters, while cross-correlation, Granger causality, and Vector Autoregression (VAR) have been …

Exogenous Variables In Time Series Forecasting During Economic Volatility: A Cross-Sector And Cross-Crisis Evaluation., Orla Kavanagh

ICT

Economic disruptions pose challenges for time series forecasting, but also offer opportunities to evaluate the value of exogenous information. This study investigates the role of external indicators during the COVID-19 pandemic and the 2025 U.S. tariff shock, using data from Dell (technology) and Boeing (aerospace/manufacturing). ARIMA, SARIMA, and LSTM models were tested with candidate exogenous variables. Results indicate that the benefit of external signals is highly context-dependent: sector, disruption type, and forecast horizon all influence effectiveness. LSTM models generally excel at longer horizons with exogenous inputs, while ARIMA-based models perform better short-term. Findings emphasize that careful variable selection, ongoing evaluation, …

Understanding Model Behaviour And Interpretability In Time Series Forecasting: A Deep Dive Into Lstm And Gru With Xai Techniques., Federico Ariton

ICT

Time-series forecasting is widely used in data analytics, yet the interpretability of deep learning models remains a key challenge. This study compares Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) models using a dual interpretability framework that combines attention mechanisms and SHAP analysis. Two feature sets were evaluated across assets with different volatility regimes: Solana and Shiba Inu (high volatility), Bitcoin (moderate volatility), and Apple (low volatility). Results show that autoregressive features achieved the lowest forecasting errors, while volatility- and momentum-based indicators provided stronger interpretability. GRU performed best in moderate-volatility conditions, whereas LSTM demonstrated more consistent performance across varying …

Traditional Vs Deep Learning Approaches For Efficient Electricity Consumption Prediction., Lucas Sant’Ana

ICT

Accurate forecasting of electricity demand is critical for reliable energy planning, resource allocation, and policy design. Traditional statistical models, such as ARIMA, SARIMA, and ARIMAX, have been widely applied but remain constrained by linear assumptions, limited temporal memory, and inflexibility in handling multiple exogenous drivers. In this study, we systematically compare these approaches with multivariate Long Short-Term Memory (LSTM) networks, which are capable of capturing nonlinear dependencies, long-term temporal dynamics, and multivariate interactions. Historical electricity consumption data were combined with weather variables, including temperature, wind speed, and rainfall, and pre-processed through cleaning, scaling, and temporal alignment. Statistical baselines and deep …

Credit Card Fraud Detection., Sonia Ndonga

ICT

This capstone project applies machine learning to detect credit card fraud, addressing a critical financial threat to banks and payment providers. Using an anonymised dataset of 284,807 transactions, which is highly imbalanced with only 0.172% fraudulent cases, three models—Logistic Regression, Random Forest, and Gradient Boosting—were developed and evaluated. The pipeline incorporates data preprocessing, feature engineering, hyperparameter tuning, cross-validation, and interpretability analysis using SHAP values, SHAPASH, and permutation importance. Random Forest achieved the highest performance with an ROC AUC of 0.97 and Average Precision of 0.66. The study also considers fairness, threshold optimisation, and practical deployment strategies, providing a robust automated …

Comparison Of Convolutional Neural Network Architectures For The Classification Of Microscopic Images: Performance Evaluation Of Lightweight Models., Cristian Ricardo Gonzalez Donoso

ICT

This study evaluates lightweight convolutional neural networks (CNNs) for multi-class classification of microscopic fungi images in CPU-only environments, addressing resource-limited laboratories and educational settings. Four pre-trained models—EfficientNetV2-B0, MobileNetV3-Small, MobileNetV2, and NASNetMobile—were compared on the DeFungi dataset (9,114 images, five fungal classes) using stratified training, realistic data augmentation, and two-stage transfer learning. Performance was measured via accuracy, macro-F1, precision, recall, and efficiency metrics including training time, inference latency, model size, and FLOPs. EfficientNetV2-B0 achieved the highest accuracy, while MobileNetV3-Small provided the best macro-F1 and lowest computational cost, making it the optimal choice for CPU-only deployment. Findings highlight the trade-off between predictive …

Enhancing Fake News And Rumor Detection Using Metadata, Context, And User Interaction With A Hybrid Gcn-Transformer., Lucas Romulo Zuin Gigli

ICT

The increasing influence of social media has significantly impacted how news spreads among users, making the detection of misinformation, including fake news and rumours, a critical task. Previous research has explored content-based and metadata-driven approaches, while recent advancements have leveraged Graph Neural Networks (GNNs) such as GCNs, GATs, and SAGE to analyse user interactions and propagation patterns. This thesis investigates the effectiveness of a Hybrid Graph-Transformer Convolutional Neural Network (GCN-Transformer) for fake news detection, combining graph-based learning with Transformer attention mechanisms. Through in-depth data analysis and advanced detection techniques, this model aims to enhance predictive performance and mitigate the spread …

Analysis Of Customer Churn In The Banking Sector And Its Mitigating Factors., Fiachra O Callaghan

ICT

This project investigates customer churn within the banking sector, analysing a dataset of 10,000 customers to identify patterns and potential drivers of attrition. By leveraging data analytics and visualization techniques, we aim to understand the demographic and behavioural factors influencing customer decisions to leave for competing banks. The study applies the CRISP-DM framework to structure the analysis, enabling iterative refinement of objectives and hypotheses as insights emerge. Findings indicate that certain customer segments, such as those aged 40–60, are more prone to churn, highlighting opportunities for targeted retention strategies, including fee reductions and tailored loyalty programmes. By integrating academic research …

Credit Card Default Prediction Using Machine Learning., Yassine Zohair

ICT

This project investigates the use of machine learning to predict credit card payment defaults, aiming to help banks and credit card companies mitigate financial losses. Using historical customer data, including demographics, income, education, and previous payment behaviour, three machine learning algorithms were implemented to forecast the likelihood of default. Techniques such as cross-validation, hyperparameter tuning, and SHAPASH were applied to improve model performance and interpretability. Accurate prediction of potential defaulters enables financial institutions to take proactive measures, such as adjusting credit limits or providing targeted financial guidance, thereby enhancing risk management and customer retention.

An Enhanced Deep Learning Framework For Crop Disease Detection Using Gan-Based Data Augmentation., Andrew Mc Guinn

ICT

Plant disease detection is a critical challenge in agriculture, where limited annotated datasets often hinder the training of robust deep learning models. This thesis explores the application of generative adversarial networks for data augmentation in the classification of six classes of sorghum diseases, addressing the scarcity of real training data. A lightweight convolutional neural network  (CNN) was employed as the classifier, trained under three experimental conditions: using only real images, a stratified mix of real and DCGAN-generated samples, and a stratified mix of real and WGAN-GP-generated samples.

The baseline CNN, trained on 279 real images, achieved an accuracy of 89.7%, …

Discovering Latent Themes: Mixed-Methods Comparative Analysis Of Topic Extraction And Clustering., Laura Byrne

ICT

This study investigates whether clustering and topic modeling can uncover themes within the 167 verses of the King James Version of the Book of Esther. A standardized preprocessing pipeline was applied, and TF-IDF and sentence-embedding feature spaces were used to evaluate topic extraction (NMF, GSDMM, BERTopic, Top2Vec) and clustering (HDBSCAN, DBSCAN, Gaussian Mixtures, Agglomerative) using coherence, cluster validity, lexical distinctiveness, and cross-model similarity metrics. NMF provided the most interpretable topics, GSDMM yielded compact low-overlap topics, BERTopic offered entity-centered groupings, and Top2Vec highlighted broad thematic regions. Cross-model analysis revealed overlapping motifs, notably decree and banquet themes, but clusters and topics did …

Leaf Classification Using Convolutional Neural Networks And Vision Transformers, Louis Wilkie

ICT

This study investigates deep learning techniques for plant image classification using four publicly available datasets: Folio, Flavia, Swedish, and Sugarcane. Three modelling approaches—Convolutional Neural Networks (CNN), Vision Transformers (ViT), and hybrid CNN-ViT ensembles—are evaluated. Models leverage PyTorch implementations and pretrained weights from the Hugging Face hub, with hyperparameter tuning applied to optimize accuracy. Transfer learning with pretrained ViT models achieved the highest performance, attaining 100% accuracy on Folio, Flavia, and Swedish datasets, and 94.07% on Sugarcane. CNN and CNN-ViT models showed comparable results, but pretrained CNN-ViTs were more resource-efficient. The findings highlight the importance of evaluating multiple pretrained architectures to …

Improving Fairness In Convolutional Neural Networks For Demographic Face Classification., Leandro Andrade

ICT

This study examines racial bias mitigation in Convolutional Neural Networks (CNNs) for demographic face classification using the FairFace dataset. Three architectures—ResNet50, VGG19, and InceptionV3—are evaluated, with dataset balancing strategies including undersampling and class weighting. Results indicate that InceptionV3 with class weighting achieves the most consistent performance across racial groups, with improved F1-scores and generalization through hyperparameter optimization and data augmentation. Challenges remain in distinguishing visually similar groups, highlighting the need for equitable datasets and fairness-aware training. These insights are critical for ensuring accuracy and fairness in applications such as law enforcement, healthcare, and human–computer interaction.

Implementation Of Time Series And Neural Networks For Forecasting Agricultural Prices In The Irish Market: A Comparative Analysis Of Milk, Beef, And Potatoes., César Augusto Núñez

ICT

Agricultural price volatility represents a central challenge for the Irish agri-food sector, affecting the stability of producers, cooperatives, and policymakers. This study aimed to compare three predictive approaches applied to strategic commodities such as milk, beef, and potatoes: a traditional statistical time series model (SARIMA) and two deep learning architectures (RNN and LSTM). Using historical price series collected over a decade, the models were developed and evaluated following a rigorous methodological process that included data preparation, algorithm training, and validation of results using performance metrics widely used in time series research. The findings show that the SARIMA model was most …

Enhancing Uk Electricity Price Forecasting Using Deep Learning., Stephen Cooke

ICT

Accurate short-term electricity price forecasting (EPF) is crucial for efficient operation of the UK’s multi-layered power market, impacting generators, traders, the ESO, and policymakers. Prices are highly volatile and non-linear due to renewables, demand fluctuations, and market coupling across Day-Ahead, Intraday, and Balancing Mechanism venues. Traditional statistical models often fail under such dynamics, while machine learning and deep learning approaches—particularly LSTM, GRU, and hybrid architectures—effectively capture temporal dependencies and exogenous drivers. Empirical evidence shows that these models outperform classical baselines, enabling more accurate scheduling, risk management, and financial savings.

Evaluating The Effectiveness Of Variational Mode Decomposition(Vmd)-Enhanced Hybrid Models For Forecasting Volatile Price And Indicator Time Series., Danny Allen

ICT

Forecasting volatile time series, such as commodity prices and market volatility indices, is critical for financial and operational decision-making. This study evaluates the impact of Variational Mode Decomposition (VMD) on hybrid forecasting models combining ARIMA, GARCH, and LSTM. Using gold, platinum, and VIX datasets, models were assessed via MAE, RMSE, and computational time. Results show that VMD improves ARIMA forecasts, and achieves optimal performance when combined with ARIMA-LSTM using performance weighting reducing MAE by 0.62-3.81 and reducing RMSE by 1.48-6.20. Conversely, VMD integration with ARIMA-GARCH and LSTM decreased accuracy. While VMD enhances predictive performance, it increases computational time substantially, increasing …

Exploring The Determinants Of Life Expectancy At Birth: Predicting And Forecasting Global Health Trends Using Statistical, Machine Learning, And Deep Learning Models., Emma Rath

ICT

Accurate life expectancy forecasting is essential for health policy planning, yet research comparing statistical, machine learning, and deep learning approaches under real-world constraints remains limited. This study evaluates ARIMA/ARIMAX, tree-based, and neural network models using Irish and global datasets, considering small samples, missing data, and COVID-19 shocks. ARIMAX with lagged socioeconomic variables outperformed LSTM and other ML/DL methods. Income-based stratification improved predictive accuracy and interpretability, with SHAP analysis highlighting GDP per capita for developed countries and school enrolment and trade indicators for developing contexts. Results provide practical guidance for policymakers and establish limits for model complexity under constrained health data.

Improving The Completeness Of Food Composition Databases Using Predictive Analysis., Carla Arenhart

ICT

This study investigates the use of machine learning regression models to impute missing micronutrient values in Food Composition Databases (FCDBs), focusing on the FAO/INFOODS dataset. A cascading prediction methodology leverages nutrient interdependencies to systematically estimate missing values. Four models—Random Forest (RF), Support Vector Machine (SVM), Gradient Boosting Machines (GBM), and Deep Neural Networks (DNN)—were evaluated using MAE, MSE, RMSE, and R². RF and GBM achieved the highest predictive accuracy for protein, phosphorus, calcium, and magnesium, demonstrating that ML-based predictive analytics can provide a more reliable alternative to traditional imputation methods. These findings support improved dietary assessments, nutritional research, and data-driven …

Demand Forecasting And Inventory Optimization In Mid-Sized Grocery Retail Using Machine Learning: A Data-Driven Approach To Minimizing Stock-Outs And Waste., Dragos Andrei Ungureanu

ICT

Mid-sized grocery retailers face a persistent challenge in balancing on-shelf availability with minimizing spoilage of perishable goods. This dissertation addresses this issue by developing a data-driven forecasting and inventory simulation framework within Microsoft Fabric, leveraging scalable data ingestion, Spark-based processing, and advanced machine learning. Using multi-year transactional data enriched with holiday schedules, promotions, and macroeconomic indicators, the study compares classical ARIMA models with XGBoost to capture complex demand patterns. Rigorous hyperparameter tuning in a distributed environment demonstrates that XGBoost outperforms baseline models in terms of MAE and MAPE, particularly during promotion-driven spikes. Inventory simulations based on these forecasts reduce stock-outs …

Comparing Custom And Transfer-Learning Cnn Models For Chest X-Ray Classification: Evaluating Performance And Scalability With Kubernetes Orchestration., Christopher Anich

ICT

Chest radiography (CXR) is the primary imaging tool for respiratory diseases, but interpretation can be time-consuming and requires expertise. This study compares a small custom CNN (Victorio) with three transfer-learning models (MobileNetV2, EfficientNet-B0, ResNet-50) for three-class CXR classification (normal, pneumonia, tuberculosis) and evaluates deployment on a Kubernetes cluster. Using a balanced dataset of ~13,500 CXRs, models were trained with ImageNet pre-training and from scratch. Transfer-learning models achieved near-perfect accuracy (99–100%) and AUROC, while the custom CNN reached 87–91%. Explainability analyses confirmed models focused on relevant pulmonary regions. Kubernetes deployment showed low latency, with most delays due to service overhead rather …

Data-Driven Decarbonisation Strategy For Residential Buildings: A Q-Learning-Based Simulation Approach., Erdenechimeg Tserendorj

ICT

This study presents an AI-driven framework for optimising energy retrofit interventions in Irish residential buildings, combining Q-Learning (QL), surrogate models (ANN, RF), and a Genetic Algorithm (GA) optimiser. Q-Learning simulated sequential retrofit decisions, generating a dataset used to train surrogate models for predicting Building Energy Rating (BER) and CO₂ outcomes. The GA explored retrofit combinations to maximise BER improvement, CO₂ reduction, and financial returns. The framework identified strategies achieving up to 928 kWh/m² BER reduction and 1,563 kg CO₂ reduction per building, with strong ROI and payback within 15 years. The modular, interpretable pipeline demonstrates the practical value of reinforcement …

Chess Evaluation And Player Profiling Using Convolutional Neural Networks (Cnns) And Spatial Recognition., Joel D’Mello

ICT

This thesis explores the feasibility of employing data analytics techniques in chess, with the purpose of profiling player styles and building a comprehensive chess analytics platform. The data set consists of over 20,000 anonymized games, and therefore, the study involved feature engineering, classification and visual analytics, in order to gain more insight into player decision making in chess. The data pre-processing part of analysis involved parsing Portable Game Notation (PGN) files, and feature engineering positional characteristics - material imbalance, pawn structure, king safety, and piece mobility - along with quantifying the sample using Average Centipawn Loss (ACPL) using Stockfish. ACPL …

Forecasting Hourly Police Call For Service Volumes: A Comparative Analysis Of Statistical, Machine Learning And Neural Network Models For Operational Planning., Patrick Duggan

ICT

This study investigates hourly call volume forecasting for New Orleans police service data, comparing statistical, tree-based, and deep learning approaches. ARIMA models represent traditional methods, XGBoost and Random Forest serve as tree-based ensembles, and Gated Recurrent Units (GRUs) provide deep learning alternatives, with a naive seasonal model as a baseline. Models are evaluated using a practical, expanding time-block framework simulating real operational deployment. Results show GRUs achieve the highest accuracy (R² = 0.74, MAPE = 13.5%), with XGBoost performing similarly, while ARIMA underperforms. Additionally, a Random Forest model offers interpretability, identifying key factors that drive forecasting performance, providing actionable insights …

Enhancing Financial Fraud Detection Using Explainable Deep Learning Models On Simulated Big Data Architectures: A Comparative Analysis With Traditional Methods., Aoife Yang

ICT

Financial fraud poses a growing global challenge, driven by the rapid expansion of digital banking, e-commerce, and mobile payments. Traditional rule-based and early machine learning systems struggle to detect novel and sophisticated fraud patterns in real time. This research investigates the integration of deep learning, explainable artificial intelligence (XAI), and big data technologies to enhance financial fraud detection. A scalable data pipeline is proposed to process large volumes of transactional data, improve detection accuracy, and provide interpretable insights for stakeholders. The study highlights the potential of combining advanced AI techniques with explainability to strengthen the transparency, effectiveness, and trustworthiness of …

Predicting S&P Corporate Credit Ratings Using Financial Ratios And Machine Learning: An Analysis Of European Non-Financial Companies., Gabriele Frattaroli

ICT

This study investigates the prediction of multi-class S&P corporate credit ratings for European non-financial firms from 2010 to 2024 using a machine learning framework grounded in financial fundamentals. To ensure robustness and generalizability, the analysis excluded the Year variable, which was identified as a source of data leakage. After this correction, non-linear ensemble models demonstrated a clear advantage over linear baselines. The top-performing Random Forest model achieved a weighted F1-score of approximately 0.60, more than doubling the performance of the Logistic Regression benchmark used as a baseline (0.26), with most misclassifications concentrated in adjacent rating categories. This indicates that while …

Exploring Workforce Attrition In Early Childcare Sector Using Data Segmentation, Machine Learning And Artificial Neural Networks., Anna Fontani-Tankovska

ICT

This research presents a framework of workforce attrition exploration based on integration of data segmentation, feature selection and classification models. K-prototypes clustering and generation-based segmentation are applied to create data subsets processed using logistic regression, random forest, naïve Byes, and artificial neural networks. By segmenting data, this study aims at enhancing models classification capacity. To address data imbalance and improve models ability to identify minority class, synthetic minority oversampling technique (SMOTE) and adaptive synthetic sampling approach (ADASYN) are applied, with average recall rate improving from 0.39 to 0.61 (SMOTE) and 0.50 (ADASYN). Feature selection using Recursive Feature Elimination with SVC …

Predicting Monthly Weather Anomalies In Ireland: A Comparative Study Of Machine Learning And Deep Learning Models., Fiona Behan

ICT

Abstract

This study investigates the application of statistical, machine learning, and deep learning methods to detect monthly weather anomalies in Ireland between 1960-2024. Climate variability is intensifying globally, increasing the urgency for accurate detection of unusual weather events. Using publicly available data from Met Eireann provided by the CSO PxStat Open Data Portal, the research applied comprehensive preprocessing including Bayesian Ridge iterative imputation, temporal and seasonal feature engineering, and ensemble statistical anomaly labelling based on z-score, interquartile range, and rolling residual analysis. Four models were developed and evaluated: Isolation Forest, XGBoost, Long Short Term Memory networks, and Gated Recurrent Units, …

Investigating The Effectiveness Of Traditional Vs Hybrid Time Series Models In Operational Planning., Cristina Priolo

ICT

This study conducts a systematic comparison of traditional and hybrid time series forecasting models for operational planning in a SaaS company, using large multichannel, multilanguage datasets aggregated at six-hour intervals. Classical models (ARIMA, SARIMA, ETS) are evaluated against hybrid models (ARIMA-ANN, SARIMA-ANN, ETS-ANN) to assess their ability to capture both linear and nonlinear patterns. Forecasts over a three-month horizon were evaluated using RMSE, MAE, and MAPE, with hyperparameter optimization applied to all models. Results show that hybrid models, particularly ARIMA-ANN and SARIMA-ANN, outperform traditional models in predicting volatile and high-volume data, while traditional models remain competitive for stable datasets. These …