ARC (Academic Research Collection) (College Dubin)
Not a member yet
359 research outputs found
Sort by
Analysing Natural Language Processing Techniques: A Comparative Study of NLTK, spaCy, BERT, and DistilBERT on Customer Query Datasets.
Get PDFThis study examines the role of sentiment analysis in customer queries, emphasising its impact on brand perception and the risks of poor query management. It compares the performance of NLP models—NLTK, spaCy, BERT, and DistilBERT—on customer query and feedback data. The findings show that BERT and DistilBERT produce similar results, often categorising queries as neutral, indicating their strength in handling diverse sentiments. NLTK and spaCy also share performance patterns. The research offers insights into the capabilities and limitations of these models in sentiment analysis
Application of Machine Learning algorithms to evaluate the changes in energy consumption in the Leinster area and subsequently the impact on consumer behaviour in the commercial sector.
Get PDFThis study focuses on predicting electricity consumption through data analytics and ensemble learning methods, addressing fluctuations influenced by external economic factors. Techniques like Gradient Boosting Regressor (GBR) and Random Forest Regressor (RFR) proved effective due to their ability to generalise well with new data. CRISP-DM served as the guiding methodology, supported by robust preprocessing techniques such as winsorisation to handle outliers, feature selection to refine variables, and scaling to standardise data for improved model performance.
The research involved datasets from non-residential clients and data centres, uncovering consumption patterns through visualisations in Tableau. Analysis showed that County Dublin and Kildare were among the highest electricity consumers in Leinster from 2015 to 2022. Advanced feature selection improved model accuracy by removing variables with low correlation to the target, while preprocessing steps like one-hot encoding and data scaling ensured optimal input for regression models.
Results highlighted the predictive strength of ensemble methods, with GBR and RFR achieving high R² scores, low RMSE, and robust cross-validation performance. GBR particularly excelled with strong reliability across data subsets and balanced training and testing accuracy. While the study reinforced existing insights into ensemble learning\u27s capabilities, it demonstrated the practicality of these models for handling tabular data and extracting actionable findings from limited datasets
Maize Crop Pests and Diseases Classification Using Hybrid Models.
Get PDFThis research focuses on improving the detection and classification of maize crop pests and diseases to enhance agricultural yield and food security. A dataset comprising 5389 images of maize conditions (healthy, pest-affected, and disease-affected) across seven classes was used. The images underwent preprocessing, including resizing to 299x299, class balancing using augmentation techniques, and noise reduction with Gaussian filtering.
Feature extraction utilised EfficientNetB0 and InceptionV3 architectures, with PCA employed for feature selection. Classification was conducted using a Support Vector Machine (SVM) with a One-vs-One strategy, alongside a baseline 2D CNN model. Data engineering included label encoding, standardisation, and an 80:10:10 train-test-validation split. Hyperparameter optimisation was performed via Grid Search CV for SVM and Random Search for the 2D CNN.
The best-performing model, EfficientNetB0+SVM (299x299), achieved an accuracy of 93%, outperforming other models, including the standalone 2D CNN, which reached 78%. This underscores the advantage of hybrid models over standalone CNNs in classification tasks for pest and disease detection in maize crops
Cross-cultural management in Dublin: an analysis of the practices and challenges faced by managerial staff in Irish businesses.
Get PDFCross-cultural studies are gaining increasing relevance in both academia and the corporate world (Arslan, 2001; Chrobot-Mason, et al., 2007; Molinsky, 2013; Meyer, 2014). Ireland, with its immigration policies and incentives for multinational companies, is attracting skilled individuals, leading to a growing diversity in the country. Local businesses frequently hire foreigners, enriching teams with diverse perspectives and creativity. According to the Irish Census (CSO, 2023), over 420,000 immigrants aged 15 and over were employed in Ireland in 2022, surpassing the number of employed Irish nationals.
Highly culturally diverse workplaces necessitate specific management practices to address challenges arising from cultural misunderstandings, unclear communication, prejudices, and low cultural awareness. Effective cross-cultural management (CCM) involves fostering a team that cooperates and integrates successfully while respecting individual uniqueness. This research aims to understand the CCM landscape in Dublin, focusing on the practices, skills, and challenges faced by managers of cross-cultural teams.
Additionally, this study seeks to provide valuable insights into CCM in Ireland, addressing the gap in locally produced scientific and work papers. Employing a qualitative approach, the research combines primary and secondary sources. A comprehensive literature review was conducted, followed by interviews with three Dublin-based managers. The data gathered were critically analyzed, leading to useful recommendations for managers of cross-cultural companies
Using Machine Learning to identify hate speech and offensive language on Twitter.
Get PDFThe central theme of this project is the application of Machine Learning to identify both hate speech and offensive language on Twitter. We chose this topic for its ethical relevance in the technological environment and its business potential. This topic raises concerns such as cyberbullying and the existence of a hostile environment for users. For this reason, we sought to implement four different models to create an automated system capable of identifying and categorizing whether specific content is offensive, non-offensive or neutral
Development And Optimisation of Convolutional Neural Networks (Cnns) to Predict the Nutrition and Sustainability Scores of Foods from Crowd Sourced Images
Get PDFThis research explores the use of Convolutional Neural Networks (CNNs) for the automated classification and profiling of food products based on publicly sourced data. With so many food products available to consumers worldwide and by default, their labels, presents a challenging task for food business operators to conform to the complex regulations and for regulators to verify compliance with these requirements. With the increasing importance of data analytics in various domains, including food chains and agriculture, this study addresses the need for efficient and accurate methods for the classification and eco / nutritional profiling of foods in a timely manner
Movie Recommendation System.
Get PDFThis project is focused on implementing a Movie Recommendation System with the use of Machine Learning. The system was developed in Python and the datasets used were \u27Movies\u27 and \u27Ratings\u27 from MovieLens 25M. This project was developed with the CRISP-DM methodology and each of the phases is detailed in a report and Jupyter Notebook.
The system is a hybrid combining best qualities of collaboration filtering and user grouping. In the project we compare some models\u27 accuracies, upgrade a chosen model and show the improved performance of our hybrid model that used the SVD algorithm. We are able to find recommended movies based on user ratings
Assessment of the Impact of Various Feature Extraction Techniques on the Effectiveness of Music Genre Classification in Neural Network Models.
Get PDFThis research focuses on Music Genre Classification (MGC) using Convolutional Neural Networks (CNNs) and various datasets, including raw audio files (WAV) and extracted features such as Mel Spectrograms (MS), Mel-Frequency Cepstral Coefficients (MFCC), and Chroma Features (CF). The study employs Explanatory Sequential Mixed Methods (ESMM), combining qualitative research and experimental analysis to explore different model inputs and their performance. Several CNN-based models, including 2D CNN, 2D CNN-LSTM, 1D CNN, and 1D CNN-LSTM, were tested. However, the models generally underperformed, with most achieving accuracy of 10% or lower, and the best model (raw audio 1D CNN) reaching only 20%. The research discusses troubleshooting, model limitations, and future recommendations, including potential reasons for the low performance compared to related works
Evaluation and Development of Innovative NLP Techniques for Query-Focused Summarization Using Retrieval Augmented Generation (RAG) and a Small Language Model (SLM) in Educational Settings
Get PDFThis research explores the development and evaluation of Query-Focused Summarization (QFS) techniques for educational content, leveraging a range of NLP models, including traditional extractive and abstractive methods, Small and Large Language Models (SLMs & LLMs). The study emphasizes the effectiveness of text preprocessing strategies, comparing original, minimally processed, and fully lemmatized text across these models. To enhance scalability and cost-efficiency, Retrieval-Augmented Generation (RAG) frameworks were applied, successfully reducing input tokens while maintaining high summarization quality.
The research involved the testing of various models on a general-purpose QFS dataset, simulating educational scenarios. Significant findings highlight the superior performance of LLMs, particularly Llama 70B and GPT-4o, though SLMs like GPT-4o mini demonstrated a strong balance between cost and performance. The integration of advanced prompt engineering and optimization techniques further improved model output.
The study also incorporated primary research with educational stakeholders, shaping a practical framework for the implementation of NLP-based summarization tools in educational environments. This framework emphasizes the importance of flexible, scalable systems that promote transparency between teachers and learners while addressing the need for ethical AI practices. The research offers a solid foundation for developing NLP-driven tools that enhance learner engagement and comprehension, laying the groundwork for further experimentation and real-world integration
ML Predictive Model for Earthquakes Integrating Mass, Distance, Gravity, and Magnitude.
Get PDFThis research investigates the application of machine learning regression models to improve earthquake prediction by integrating geophysical and astronomical factors such as Earth-Moon gravitational forces, their varying distances, and localized gravity fluctuations. Using data from 2011 to 2024, sourced from the US Geological Survey (USGS) and web scraping, the study tested models across four dataset proportions (25%, 50%, 80%, and 100%) with a 70:30 train-test split. The XGBRegressor model emerged as the best performer, achieving an R² score of 0.8706 on training data and 0.8632 on test data, along with a Mean Squared Error (MSE) of 0.1114 and Mean Absolute Error (MAE) of 0.2473. Feature importance analysis via Information Gain, ANOVA, and Lasso identified gravity variations as significant predictors, while Moon-Earth distance showed no statistically significant impact. Despite promising results, the moderate R² score highlights the model\u27s limitations in capturing the full complexity of earthquake prediction. Future research should incorporate more geological metrics, real-time seismic data, and advanced machine learning techniques, such as convolutional and recurrent neural networks, to enhance predictive capabilities