8333 research outputs found
Sort by
An investigation of the Impact of Instagram Advertising on the Decision-Making Process of International Students Considering Third-Level Education in Dublin
No full textThis study investigates how Instagram ads affect international students' choices considering going to college in Dublin. As digital communication changes the way people choose to learn, Instagram, which is visual and engaging, is a vital tool for drawing potential students. The main goals of this study were to look into how digital communication with schools affects students, how Instagram helps bring international students to Dublin, how Instagram ads affect international students' choices of university and course, and how much international students interact with Instagram. The study uses conversations with potential students to show that Instagram significantly impacts student decisions through visual material, interaction with school profiles, and personal recommendations. These results show how vital strategic Instagram marketing is for educational organizations and give valuable tips for making digital marketing plans more effective. This study helps us understand how the role of social media is changing in marketing for foreign education and gives schools a way to improve their outreach.
This study investigates how Instagram ads affect international students' choices considering going to college in Dublin. As digital communication changes the way people choose to learn, Instagram, which is visual and engaging, is a vital tool for drawing potential students. The main goals of this study were to look into how digital communication with schools affects students, how Instagram helps bring international students to Dublin, how Instagram ads affect international students' choices of university and course, and how much international students interact with Instagram. The study uses conversations with potential students to show that Instagram significantly impacts student decisions through visual material, interaction with school profiles, and personal recommendations. These results show how vital strategic Instagram marketing is for educational organizations and give valuable tips for making digital marketing plans more effective. This study helps us understand how the role of social media is changing in marketing for foreign education and gives schools a way to improve their outreach
Report for Optimizing Renewable Energy Management through Solar Power Forecasting
No full textWith the shift towards renewable energy has increased exponentially, improving the grid stability, encouraging the transition navigating towards renewable energy sources and an efficiently predicting model for solar power are important for enhancing the management of renewable energy. This research study aims to use deep learning base approach specifically, Transformers for learning the complex patterns and trends in solar time series data. We also compared transformers based approach with various machine learning approaches like the Decision Tree Regression model, KNN (K-Nearest Neighbors) Regression model, Gradient Boosting Regression model. Transformer model has shown exceptional performance compared to other models, where the values of R², MSE, and MAE are 0.99999, 0.97, and 0.56 respectively while other machine learning approaches also performed better. Transformer model able to make better predictions because it able to learn the long range dependencies of the solar data as it is time series data. The transformer model has shown significant enhancements in terms of identifying dependencies and complex patterns present in the data, this helps in improved predicting accuracy. This research study aims is understand and show the capabilities of enhanced deep learning and machine learning models for improving the predicting accuracy of solar power. The knowledge is acquired to contribute enhancements in grid integration and energy storage management. Future works can concentrate on working with real-time data to improve prediction accuracy and develop more integrated models. This study contributes to reliable and sustainable energy systems
Predictive Modelling for Power Consumption in Tetouan, Morocco Using Machine Leaning Method
No full textAs the world’s population continues to grow, the demand for electricity consumption is on the rise, necessitating accurate prediction to meet increasing energy needs. This research uses machine learning techniques to predict power consumption across three zones in Tetouan, Morocco, a city facing fast Urbanization and increasing demands for energy. Accurate power consumption in this region are important for good energy management and planning to help reduce cost and ensure consistent power supply. Focusing on four machine learning models, Decision Trees, Random Forests, Long Short-Term Memory (LSTM) networks, and XGBoost and using historical dataset comprising DateTime, Temperature, Humidity, Wind Speed, General Diffuse Flow, Diffuse Flows, and zone-specific power consumption variables, from the UCI Machine Learning Repository, the objective of this study is to evaluate the predictive accuracy of the these models, identify main predictors of power consumption and assess their computational efficiency. The findings indicate that the XGBoost model provides the highest predictive accuracy followed by the Random Forest model. The LSTM model, effectively capture temporal dependence’s, making it good for sequential predicting. The Decision Tree model serves as baseline with lower performance compare to the other models.
This research contributes to the field of energy management and demonstrates the effectiveness of advanced predictive modeling techniques narrowed to the unique characteristics of Tetouan. The knowledge gained can help in optimizing energy distribution, reducing cost and promoting sustainable development initiatives in Tetouan’s region
Enhancing Next-Day Stock Price Prediction Accuracy and Reliability: A Comparative Study of Bi-GRU, Transformer, and Hybrid Models
No full textThis research project is an attempt at enhancing the accuracy and performance of nextday stock price predictions using various machine learning models. In this paper, we look at how well Linear Regression, Bidirectional Gated Recurrent Unit, Neural Networks, and Transformer models do against stocks taken from Google historical data from January 2015 to December 2023. We did extensive experimentation and found out that a Linear Regression model was actually the most fitting one when compared to the large capacity models such as Bi-GRU and Transformer. Evaluations were based on the mean squared error and mean absolute error. The results indicated that, on the one hand, more complex models do have the ability to capture non-linear trends very well, but on the other hand, less complicated models perform much better if there is a strong presence of linearity in data. This research contributes to the current body of literature and continuing research on financial forecasting since it gives insights into when some of these different machine-learning approaches have relative strengths and not only that—this work also emphasises model selection based on the characteristic of data
SumBot: An enhanced multilingual Document Summarization using LLMs
No full textIn a time where knowledge is available in excess, both written and spoken, summarising is an especially useful ability. Long texts are condensed into clear, comprehensive formats by summarization, which facilitates efficient communication and decision-making. This problem is addressed by automated document summarising, which uses Large Language Models (LLMs) and Natural Language Processing (NLP) to extract pertinent information from texts. Using extractive or abstractive approaches, this procedure identifies important words or concepts, preserving the main ideas of a document while eliminating unnecessary details. A unique hybrid framework called SumBot was created especially for the field of scientific literature to facilitate multidocument scientific summarization (MDSS). To produce high-quality summaries, this framework makes use of several Sentence Transformers and models from the T5 family. To adequately summarise entire material, the research focuses on analysing various kinds of LLMs and considering diverse document styles and languages. The study intends to improve automated summarization's accuracy and efficiency by analysing these models' performance, making it a useful tool for managing massive amounts of data in a variety of scenarios. This method helps better decision-making processes in a variety of disciplines and enhances information retrieval
Stock Market Prediction Using Financial News Sentiments and Technical Indicator Data with Machine Learning Models and LIME for Explainable Insights
No full textThe stock market is comprised of many input factors, which include stock prices, news sentiments, and technical ones. The general problem of how to forecast stock prices has not been fully solved due to the constant fluctuations of shares. This research aims to enhance stock price forecasting based on the utilization of heterogeneous data and machine learning algorithms and make the black box prediction more interpretable for investors. The purpose is to improve the forecast precision and gain a deeper understanding of the opportunities for the markets. The study involves the quantitative data of the stock prices such as open, high, low, closing, volume, SMA (Simple Moving Average), EMA (Exponential Moving Average), RSI (Relative Strength Index), BBANDS (Bollinger Bands), and News sentiment score data. The methodology has dwelled into three widely used models RandomForestRegressor, SVR (Support Vector Regressor), LSTM (Long-short term memory) and their hyper-parameter tuned versions for better results. Among all the models, the fine-tuned SVR model has outperformed others. The fine-tined SVR model achieved an MSE (Mean Square Error) of 0.518 and an MAE (Mean Absolute Error) of 0.566. The integration of technical indicators and news sentiment scores along with the LIME (Local Interpretable Model-Agnostic Explanation) explanation can significantly benefit traders and financial analysts by providing more accurate predictions and explanations. The real-time data processing and potential biases in sentiment analysis are the challenges that can be explored further. The final objective of this research is to empower traders and financial analysts to make sound and data-backed decisions in the stock market
Optimising Gated Recurrent Unit for Intrusion Detection in Internet of Things Networks: A Comparative Analysis with Other Deep Learning-Based Methods
No full textThe research project’s primary goal is to protect the Internet of Things devices from the cyber-attacks using the advanced neural networks. This paper was discussed about the increase in accuracy as well as effectiveness of the intrusion detection in Internet of Things (IoT) network by optimising the design and hyperparameter of Gated recurrent units (GRUs). This research work was fully dedicated to the growth of an attack classifier, which was the intrusion detection system. The rapidly increasing in the number of cyber-attacks have made intrusion detection prediction research essential. It is important to maintain the security as well as integrity of the Internet of Things (IoT). Even though for predicting the intrusions there are many ways to solve this problem but the most efficient way we used in this research to solve this is Gated Recurrent Units (GRUs) which is a gating mechanism in deep learning. In this deep learning technique, we have used a specifically used the Bidirectional GRU and convolutional gated recurrent unit (ConvGRU). The different network traffic data in the existing RT-IoT dataset, we evaluated the models based on different type of attack scenario. The gated recurrent unit’s models were trained in this research and evaluated the model’s accuracy using the IoT dataset. A deep learning model based on Gated Recurrent Units (GRUs) is explained and showed results can predict future alert probabilities from an attacking source. A comparison was evaluated using evaluation metrics and from the analysis the Convolution gated recurrent unit performed well than the other two models and detect the IoT attacks with a precision of 98%, F1 Score of 98.3% and an accuracy of 98% with minimum loss value of 0.04 respectively. Based on the evaluation, using the deep learning methods, we can conclude that the intrusion detection in the IoT can be solved successfully. This entire research handled with the ethical issues which include data confidentiality as well as reliability
Machine Learning for Feature Extraction and Classification of English-language Accents in Ireland
No full textPronunciation in the English language in Ireland is of strong interest to the Irish public and the research community, both as a marker of identity and in considering interactions with modern automated speech processing tools. Qualitative linguistic research has consistently shown that the pronunciation within Ireland of the English language has shown significant differences by geographic origin and between Irish mother-tongue speakers and accents in India, Australia and elsewhere. Recent data analysis reinforces the distinctness of Irish-English from forms of English spoken in mainland Britain. Using speech samples from the wide survey published in Hickey’s (2004) ‘Sound Atlas of Irish English’, we attempt to build models to classify the Belfast and Dublin regional accents of Irish English using logistic regression, neural networks, convolutional neural network and large audio models. Evaluation by accuracy, confusion matrix and ROC curve methods showed strong classification ability for these regression and neural network models. However, performance using recent transformer-based large audio models was poor. Overall, this research points to continued future data-gathering and more modelling work while preserving privacy as promising avenues for future research, leading to greater socio-linguistic self-understanding and to reduced bias impacting consumers in Ireland
A systematic evaluation of vision transformers for galaxy classification
No full textThis study explores the effectiveness of Vision Transformers (ViTs) in the morphological classification of galaxies. This research utilizes the Galaxy10 Decals dataset for the deep learning tasks. The research focuses on three advanced transformer-based models—ViT Base, Swin Transformer, and DeiT Transformer alongside the conventional ResNet50 model. The Galaxy10 dataset comprises of 10 galaxy classes, serves as the benchmark for evaluating model performance. The ViT Base model is fine-tuned on the Galaxy10 dataset with weights pre-trained on ImageNet. The model demonstrated a robust performance due to its ability to capture complex relationships through multiple layers of multi-head self-attention. Similarly, the Swin Transformer is known for its hierarchical design and shifting windows, and the DeiT Transformer is enhanced with data efficiency techniques and knowledge distillation. Both the models showcased significant accuracy and precision in galaxy classification tasks. Evaluation metrics were included in this research such as precision, recall, accuracy, and F1 score. The metrics ensured a comprehensive assessment of model performances. The results indicates that the ViT Base model achieved the highest accuracy; however, a baseline CNN model performed faster. This research highlights the trade-off of Vision Transformers in the domain of astronomical image classification. It offers insights into their capability for detailed morphological analysis of images. The findings suggest that ViTs could be used as a general-purpose image classification technique, showing slightly better accuracy than ResNet50. Overall, vision transformers show superior ability to model contextual information and are promising tools for image classification
Enhancing the Early Detection of Dental problem through Transfer Learning Techniques in Dental Radiography
No full textThe poorest showing is represented by the sphere of public health with its focus on the most vital human health procedures and phenomena one’s teeth. Such measurements allow dental panoramic radiography to be widely accepted among dentists in diagnosing and researching such diseases and with general exposure of the whole oral area use protective measures of low radiation dose and radiation time. To detect these matters, dentists have varieties of radiography like the panoramic views which do not take much time and have a low degree of radiation and perhaps, provide the visualize of the whole area in the mouth. It may take several hours and the whole process may tend to be tiresome especially when the veterinarian is interpreting the radiographs. Modern trends have allowed the dentists to complete the analyses in a shorter time with the support of various forms of Artificial Intelligence. It is rather sad today to such a group of people who require these services to part with an enormous amount of money to be treated by a doctor or maybe even get an x-ray. Large movies are when the malformations cannot be touched with the help of fingers In such a way, big movies are utilized by doctors. It assists in the management of the modern diseases such as caries, deep caries, impacted teeth and periapical lesions; hence development of cheap dental health care services