ARC (Academic Research Collection) (College Dubin)
Not a member yet
359 research outputs found
Sort by
Detecting Fake News Using AI
No full textThis is a document that presents a strategic analysis of the project “Detecting Fake News using AI”. Developed as part of the BSc (Hons) in Computing in IT at CCT College Dublin. The goal is to analyse the potential advantages, exploiting the viability and the impact of applying Artificial intelligence to check, verify and alert about misinformation found and to answer the question “How can Artificial Intelligence be leveraged to accurately detect and combat fake news while ensuring data privacy and compliance with regulations?” and “To what extent can AI-driven misinformation detection help reduce the spread of fake news on social media platforms?”. Always based on its legality and ethicality, focused on users\u27 data protection, this project will help society protect itself from mass manipulation, avoiding, for example, influencing elections, financial markets and the public perception of reality
Comparative Evaluation of AI-Generated Synthetic Data and Real-World Data Performance in Predictive Analytics
No full textThere are growing restraints when it comes to Real World Data (RWD), these include topics such as privacy regulations, ethical concerns, and the cost of collecting the data, and they have drove an interest in AI-generated synthetic data as a potential alternative in predictive analytics. This project examines the possibilities of synthetic data and if it can act as a reliable substitute for RWD in predictive modelling. This project uses Wasserstein Generative Adversarial Networks with Gradient Penalty (WGAN-GP) to generate synthetic reproductive health data and evaluates its predictive performance against RWD using linear regression and key metrics, including Mean Absolute Error (MAE), Mean Squared Error (MSE), and R².
By completing a comparative analysis, this research assesses the viability of synthetic data in predicative forecasting, drawing on both its strengths and limitations. The findings show that synthetic data that is created by using machine learning, deep learning, and statistical validation can achieve predictive performance (97.8%) comparable to RWD (99.7%) under certain conditions (such as number of epochs and number of samples), offering an avenue for enhancing data accessibility while mitigating privacy risks. This research also investigates mode collapse, and parameters that can help mitigate against that, with an R2 result of 10% being increased to 97.8% by altering the WGAN-GP generator.
By integrating privacy evaluations and ethical considerations, this study contributes to the ongoing discourse on synthetic data applications, offering insights into its role in advancing data-driven decision-making across healthcare
Development of a Deep Learning Model for Synthetic vs. Real Image Classification Synthetic vs. Real image classification-Poster
No full textThis project develops a deep learning model to classify images as either AI-generated or real, addressing the growing challenge of synthetic media detection. Using the DeepGuardDB dataset and guided by the CRISP-DM methodology, we implemented and compared three Convolutional Neural Networks (CNNs) architectures via transfer learning. The best-performing model was further optimised using hyperparameter tuning and fine-tuning techniques The resulting model achieved strong accuracy and generalisation, making it a promising candidate for real-time deployment and practical use across diverse industries
Assemble the ensemble: A multi model approach for customer churn prediction in the gambling industry.
No full textChurn rates are remarkably high in the gambling industry, an extremely competitive landscape coupled with a severe lack of brand loyalty among its customer base makes churn prediction one of the main problems an operator will face. This paper explores the range of possible modelling solutions with a key emphasis on ensemble learning to improve on existing methods. During this exploration, a host of modelling techniques are formulated with a focus on scalability facilitated by Apache Spark distributed computing language. Thirteen variations of models, including single classifiers and ensemble families are evaluated as to their suitability in solving the problem. The limitation of the data set provided is that it is not diverse enough to encapsulate the true dynamism relationship between the customer and an operator. Nevertheless, the paper can provide a host of solutions that satisfy the goal of scalability. The project provides three separate recommendations that are flexible based on the firm needs recording ensemble classification precision scores of 86%, 87% and 95% respectively. The author proves that ensemble learning is a stronger predictive solution in the context of churn prediction in the gambling industry. In addition to demonstrating the power of ensemble learning the paper provides an application based on the author’s strongest modelling approach that is applied on a unseen validation set. The output of the application returns a list of customer account numbers who are predicted churners that internal CRM teams can use to improve processes
Dogs Emotion System
No full textFor our capstone project, we built a machine learning model that can look at pictures of dogs and figure out how they’re feeling, like if they’re happy, sad, or just chill. The idea came from how important pets are in people’s lives these days and how cool it would be to actually understand their emotions better using tech. This system will allow users to upload images of dogs, which are then analysed by a trained model to classify the dog\u27s emotional states such as happy, sad, or neutral. We followed the CRISP-DM process to build it, which basically means we went through steps like understanding the goal, getting and cleaning data, training and testing the model, and thinking about how it could be used in real life. So the proposed fully trained model can be used in many ways. It can be used in the backend of a ‘Dog classification website’ or it can be a mobile application. It could help pet owners take better care of their pets, and maybe even help catch signs of stress or sadness early. This report is about how models are trained, what is used and how It shows how tech can actually help us connect better with our pets and understand them in a smarter way
Strategic Analysis of Employment Permit Statistics and Predictive Analytics for Workforce planning in Ireland- Poster
No full textThis project analyses employment permit trends in Ireland from 2020 to 2025. It aims to help recruitment agencies and job seekers with data driven insights to enhance hiring placement. Forecasting permit demand by sector to help improve workforce planning and policy decisions
Player Transfer Market in European Football Using Machine Learning to analyse the evolution of the European football.
No full textThe study uses machine learning to analyse the European football transfer market from 2015 to 2025, revealing patterns in transfer fees and market values influenced by player attributes, highlighting the potential of data-driven insights
Predicting early hospital readmissions for diabetic patients using machine learning- Poster
No full textThis project applies machine learning to predict whether diabetic patients will be readmitted to a hospital within 30 days of discharge. Early readmissions are a costly and critical issue in healthcare, often signalling gaps in post-discharge care and risk management. Diabetic patients face unfair high readmission rates compared to the general population. According to the CDC Diabetes Report Card 37.3 million people in the U.S. or 11.3% of the population had diabetes as of 2019 (CDC, 2021). Our goal here is to develop a binary classification model capable of flagging high risk patient (\u3c 30-day readmission) based on their clinical, demographic, and administrative data. This lets healthcare institutions to take measures
Brain Tumor Classification using Deep Learning Custom CNN vs. ResNet50.
No full textThe project presents a deep learning solution to classify brain tumors through MRI images. Following the CRISP-DM framework, two Convolutional Neural Network (CNN) models were developed and evaluated, a custom CNN designed from scratch and a pretrained ResNet50 that was transfer learned and fine-tuned. Both models were assessed using standard performance metrics such as accuracy, precision, recall and F1-score. Despite the higher test accuracy achieved by the custom CNN, further interpretability indicated inconsistent attention to the actual tumor regions also known as shortcut learning. On the other hand, ResNet50 showed more reliable and clinically relevant focus which supported its selection as the final model. The selected model was deployed using Gradio, to demonstrate a real-world application with real-time predictions and visual explanation to reduce the black box nature of AI. The results showcase that despite the good performance of custom and tailored CNN models on specific datasets, generalization and real-world relevance are equally important for reliable deployment
Using Machine Learning to Predict Credit Card Fraud
No full textCredit card fraud poses a significant challenge to financial institutions, leading to substantial financial losses and declining customer trust. This project develops and evaluates machine learning models to detect fraudulent credit card transactions using a large, realistic synthetic dataset. Following data preprocessing, exploratory analysis, and class-balancing using SMOTE, four supervised models—Logistic Regression, Decision Tree, Random Forest, and XGBoost—were trained and compared. Performance was assessed using metrics suited to imbalanced classification, including AUC, Recall, Precision, F1-score, and Average Precision. Results show that XGBoost, particularly after hyperparameter optimisation, delivered the strongest performance (AUC 0.99, Recall 0.83, AP 0.70), outperforming other models and demonstrating high effectiveness in identifying fraudulent activity with manageable false-positive rates. Findings confirm that tree-based ensemble methods are well-suited to fraud detection and that appropriate resampling and tuning strategies significantly improve predictive accuracy