Information & Communications Technology

Permanent URI for this collection


Recent Submissions

Now showing 1 - 5 of 162
  • Item
    Assessing Commercial Wearables in Predicting Physical Activity: A Case Study of Apple and Fitbit
    (Dublin Business School, 2024) Shadare, Ayodeji; Adeagbo, Bolu
    This study explored the predictive accuracy of Apple Watch and Fitbit in tracking physical activity, employing advanced machine learning models, feature engineering, ensemble learning, and interpretable machine learning. Using Microsoft Power BI, an interactive dashboard was also constructed to analyze user demographics, physiological metrics, and activity patterns. The machine learning models, including Logistic Regression, Naïve Bayes, Decision Tree, Random Forest, LightGBM, XGBoost, CatBoost, and Artificial Neural Network (ANN) were comprehensively evaluated using multiple metrics. Interpretability was enhanced through Shapley values, unravelling the contribution of features to classification results. Stacking models reveal insights into their performance compared to individual models. The result generally showed that the single LightGBM model was better compared to other models and stacking. Furthermore, the dashboard insights provide a detailed exploration of user engagement across different activities, revealing variations in heart rates, distances, and calorie expenditure. This study contributes unique insights into wearable technology.
  • Item
    Competitive Analysis of Embedding Models in Retrieval-Augmented Generation for Indian Motor Vehicle Law Chat Bots
    (Dublin Business School, 2024) Mohanan, Monisha; Hoare, Terri
    This study evaluates eight embedding models in Retrieval-Augmented Generation (RAG) systems for a chatbot tailored to Indian Motor Vehicle Law. The models examined are OpenAIEmbeddings, UAE-Large-V1, all-MiniLM-L6-v2, all-distilroberta-v1, all-mpnet-base-v2, bge-large-en-v1.5, ember-v1, and gte-large. Through Cosine Similarity and ROUGE metrics, the analysis distinguishes OpenAIEmbeddings and gte-large for their superior semantic understanding. These models showed remarkable alignment with expert-generated answers, indicating their efficacy in AI-driven legal assistance. The study's outcomes underscore the importance of embedding model selection in legal chatbot development, focusing on semantic comprehension capabilities. This research is pivotal for enhancing AI legal assistance, offering insights into the effective integration of embedding models in legal technology applications.
  • Item
    Real Time Facial Expression Recognition using EfficientNetB7 Model Architecture
    (Dublin Business School, 2024) Gattu, Anwesh; Kshirsagar, Vivek
    Facial expression recognition (FER) represents a pivotal aspect of human interaction and emotional communication. This research delves into the development and implementation of a real-time facial expression recognition system using the FER2013 dataset obtained from Kaggle. Leveraging the cutting-edge EfficientNetB7 model, the study aimed to create a robust and accurate model capable of detecting various facial expressions in real-time through webcam feeds. Employing a combination of Machine Learning and Deep Learning methodologies, the project focused on training the model to categorize facial expressions into distinct emotions. The methodologies encompassed preprocessing steps, data augmentation techniques, and the design of an EfficientNet7 architecture tailored to the nuances of facial feature recognition. The model's performance was evaluated using metrics such as accuracy, precision, recall, and loss. This research encapsulates the comprehensive approach and findings achieved in the pursuit of advancing real-time facial expression recognition technology.
  • Item
    AI-Driven Spend Analysis Application: Integrating Purchase Order Classification Proactive Procurement Forecasting & Spend Visibility
    (Dublin Business School, 2024) Mehdi Rizvi, Mohd; Ezenwa Nwankire, Charles
    This research introduces a groundbreaking advancement in the realm of spend analytics within procurement, presenting an automated solution that integrates cutting-edge machine learning models with Microsoft Power BI. Utilizing Convolutional Neural Networks (CNN) for precise text classification of Purchase Orders (PO) and employing RandomForestClassifier, RandomForestRegressor, XGBClassifier, and XGBRegressor for forecasting both spend and the most procured categories, this methodology constitutes a substantial contribution. The implementation of batch file automation streamlines all process components with a single click. The CNN model enhances efficiency and accuracy by automating the classification of purchase order text, significantly reducing manual efforts in procurement. Simultaneously, the RandomForest and XGBoost models contribute to robust forecasting, delivering proactive insights. The study meticulously details the development, training, and seamless integration of these models within the Power BI environment, offering insights into both challenges and successes. Real-world application and rigorous testing validate the practicality of the solution, demonstrating improved accuracy in text-based purchase order classification and resilient forecasting capabilities. Results indicate a CNN accuracy of 70% for transaction categories, highlighting its adaptability. Forecasting models, particularly XGBoost, exhibit superior accuracy with minimal deviation, achieving a variance of 1.5%. This automated approach transforms spend analysis methodologies. The paper concludes by discussing broader implications, and potential advancements, and suggesting future avenues for refining ensemble machine-learning applications.
  • Item
    Water Quality Analysis Using Machine Learning
    (Dublin Business School, 2024) Vilas Khaire, Neelam; Abgaz, Yalemisew
    Water quality evaluation is crucial in environmental management, and utilising machine learning models improves the accuracy of predictions. This study aims to compare different machine learning models for predicting water quality before and after the monsoon season in Telangana. The dataset used in this research was obtained from the Telangana Ground Water Department. The chosen models, namely Random Forest Classifier (RFC), Support Vector Classifier (SVC), Multi-layer Perceptron (MLP), Stochastic Gradient Descent (SGD), and KNeighborsClassifier, are assessed with a specific focus on imbalanced data using Principal Component Analysis (PCA) as the model was giving perfect score due to being imbalanced which was incomparable and incorrect. The effectiveness of the models is evaluated by employing essential performance metrics, including recall, precision, and F1 score as the accuracy does not work well with imbalanced data. The pre-monsoon results indicate that RFC performs exceptionally well, with a recall of 0.988 and precision of 0.900. The monsoon transition has had a noticeable effect on RFC, as it continues to perform exceptionally well in the post-monsoon period, with an improved recall rate of 0.996 and precision rate of 0.993. SVC, SGD and MLP demonstrate consistent and strong performance in both time periods, demonstrating their ability to adapt. Notably, the KNeighborsClassifier demonstrates enhancement after the monsoon season, highlighting its sensitivity to seasonal changes. The analysis of seasonal variations was performed with help of T-test on the machine learning model performance’s. RFC demonstrates consistent excellence. The comparative analysis enhances the scientific comprehension of machine learning models in predicting water quality, providing practical implications for environmental scientists, policymakers, and stakeholders involved in water resource management.