8333 research outputs found
Sort by
Understanding the Impact of Social Media Sentiment on Financial Decision-making within the Stock Market: A Deep Learning Computational Analysis
No full textThis study considers an innovative way to explore social media sentiment analysis for stock market prediction through the use of Generative Adversarial Networks (GANs) to improve the accuracy of the forecasting models. Traditional financial theories like the Efficient Market Hypothesis (EMH), and the Random Walk (RW) theory have often overlooked the psychological and behavioral aspects of market dynamics that drive stock prices. Thus, our study incorporates the psychological component through sentiment data expressed in X (formerly known as Twitter) by designing three predictive models; Long Short-Term Memory, Random Forest, and GAN. These models were subsequently evaluated against Tesla (TSLA) and Amazon (AMZN) stock data, focusing on some major performance metrics such as accuracy, precision, and recall. In this respect, the GAN model demonstrated superior performance with an accuracy of 82.67%, precision of 70.21%, and recall of 81.11% for TSLA, and accuracy of 84.21%, precision of 85.71%, and recall of 75.00% for AMZN. In comparison, the LSTM model achieved an accuracy of 66.67% for TSLA and 53.84% for AMZN, while the RF model achieved 56.86% for TSLA and 54.00% for AMZN. This research not only contributes to the evolution of computational finance but also accentuates how decisive behavioral economics can be in understanding and predicting market trends. The results subsequently indicate that incorporating social media sentiment increases substantially the predictive power of financial models, therefore offering a more nuanced approach toward market analysis
Advancements and challenges in Autonomous Vehicles: Accelerating or braking based on Image processing
No full textTo achieve a correct decision process in autonomous vehicles and increase their safety, one of the main processes involved is the decision the vehicle must make between accelerating and braking. Deciding this will involve information the sensors retrieve from the environment and processing it using precise computing systems. However, the vehicle must incorporate trained data that involves different situations in which the vehicle must decide. The trained data must be precise and must be tested before being applied in autonomous vehicles, making sure it is precise and reliable. In this study, the investigation focuses on the effectiveness of image processing techniques, in this case, Support Vector Machine (SVM), to make the appropriate decision to accelerate or brake
Ethical Considerations in Explainable Artificial Intelligence: Transparency and Accountability in AI Decision-Making
No full textIn many industries, AI has emerged as a powerful tool, which has raised valid concerns over the opaqueness of algorithms used in decision making processes. This paper aims to investigate the ethical considerations surrounding AXI with a focus on critical industries including; healthcare, banking and criminal justice. The main question is focused on how to effectively address the problem of realizing the applications of AI techniques as well as the demands for the interpretability, explainability, and audibility of AI decisions while promoting justice, accountability, and privacy. In a similar manner, XAI approaches are assessed, prototyped within this study, and case studies were conducted to explain how XAI could be implemented. It also entails consultations with the stakeholders in order to identify some of the issues and goals that they may have regarding the use of AI in activities such as transparency and accountability. The study’s implications indicate that XAI has the potential to improve AI governance to become more transparent and fair in applying AI technologies through eliminating risks and algorithm bias, as well as strengthening the level of trust of all interested parties. Recognizing the lack of congruency between technology adoption and its ethical implications in literature, this research will help in the progression of proper implementation policies of AI solutions
Enhancing Microservice Performance: A Hybrid Model Combining Service Discovery and Circuit Breaker Patterns in Microservice Deployments
No full textMicroservices are a medium to contribute and build a dynamic, high functioning application that can be used worldwide by leveraging cloud services. The requirements can be easily integrated under different frameworks and the configurations that come with them. The main question that arises is what happens when any service of the application faces any issue or downtime. The application should be made flexible and scalable while keeping the factor of high availability in mind. With many organizations migrating to cloud-based infrastructures, it becomes complex to manage the response time and latency of the independent microservices. For instance, Amazon faces costs of 1% sales for every 100ms of delay van Vessum (2024). Therefore, enhancing the responsiveness of microservice dynamically, this paper explores two widely used design patterns: Service Discovery and Circuit Breaker. We conducted experiments using both patterns individually as well as a combined architecture on backed services developed using the Spring Boot Framework. Under the combination, it is observed that the Circuit Breaker pattern’s fault tolerance and resilience while service discovery’s efficiency for load balancing and dynamic routing brings together high performance architecture. The findings observed through experiments say that each pattern improves latency independently, while we proposed a hybrid model that significantly improves the performance by fourfold by harnessing the strengths of both patterns. This improvement contributes to the performance achieved that addresses the real issue of microservice latency, improving response time and making application fail safe
Forecasting of climatic influence on energy generation from renewable resources in Spain using Neural Network Models
No full textThe technological advancement and population growth has brought significant impact in the need for electricity throughout the world. The soar in energy demand brought significant impact on environment due to the over usage of conventional methods such as fossil fuels. The combustion of natural gases, coal and oil is releasing considerable amount of CO2 into the atmosphere which highly dangerous to our planet. This can be solved by switching from this traditional method to renewable resources such as solar, wind, biomass etc. which are naturally available in nature in sufficient amount. But their dependency on the weather is a challenging factor. This study is addressing this issue by forecasting the energy generation from renewable resources under the influence of weather. For conducting this, the study employs the advanced neural network models such as LSTM, Stacked LSTM, LSTM-CNN and LSTNet to forecast the impact of weather on renewable energy production. The comparison of results shows, LSTNet’s superior performance with 98.99% accuracy over the other models, which is evaluated based on the evaluation metrics such as mean squared error, means absolute error, root mean squared error and r- squared value. The outcomes emphasize the importance of the selection of appropriate predicting models and also showcase LSTNet as a valuable tool for future energy predictions
A Federated Learning Service Ecosystem for Secure and Flexible Model Sharing in Multi-Cloud Environments
No full textThis research work proposes a relatively new serverless federated learning (FL) model-sharing mechanism for securing flexible multi-cloud environments. The research presents solutions to important issues in decentralized machine learning such as securing fine-grained access control and privacy-preserving aggregation. This proposed framework scales across multiple AWS accounts for secure data and model sharing using a decentralized architecture that resembles the federated learning process. It uses attributebased encryption (ABE) enabled by encapsulating the attributes into Advanced Encryption Standard (AES) to allow peer-to-peer model sharing across different AWS accounts without the need for central authorities. A masked-ring protocol is implemented for decentralized model aggregation to protect user privacy during training. We have implemented and evaluated our architecture on AWS Lambda to prove that it can be used on real-world serverless platforms. The framework is highly effective and scalable as demonstrated by experiments on multiple cloud accounts by training convolutional neural networks (CNN) on a subset of the MNIST datasets for local training and local model generation. The experimental framework is designed using AWS Lambda functions to distribute the dataset across clouds, begin local model training, encrypt, and use local masks before saving to S3 buckets. The host function can then be triggered to access these local models to perform masked ring aggregation for unmasking and decrypting the aggregated model. This research is part of the work that will advance privacy-encapsulated collaboration learning in multi-cloud, keeping focused on the pragmatic balance between security resourcing vs flexibility alongside performance
Securing Containerized Environments: Implementing Role-Based Access Control with Google Kubernetes Engine
No full textThis project focuses on implementing a dynamic Role-Based Access Control (RBAC) policy within a Kubernetes Autopilot cluster to optimize resource management and application stability. The primary objective was to dynamically adjust user roles based on real-time CPU usage metrics. When CPU usage exceeded 1000 millicores, roles were elevated to "Owner" to handle increased resource demands, while roles were reverted to their original settings when CPU usage fell below the threshold. The implementation does include setting up a Google Cloud Platform (GCP) project by configuring APIs by creating a Kubernetes cluster and deploying a Pub/Sub topic for alert notifications. Open Policy Agent (OPA) was used to use dynamic RBAC policies. Automation scripts have been created to adjust roles based on alerts. The performance of the dynamic RBAC policy was evaluated by monitoring role adjustments and system performance. Results have been shown successful role escalations and reversion by showing improved resource management and application stability. Compared to traditional static RBAC the dynamic approach given better resource optimization by increasing security and operational performance with good improvements in response times and error rates. This approach has given a scalable solution for managing roles in cloud-native environments by securing optimal performance and security
Systematic Needs Analysis of Advanced Digital Skills for Postgraduate Computing Education: The DIGITAL4Business Case
No full textA 15-partner funded project, DIGITAL4Business aims to revolutionise the digital landscape in Europe, fostering strong industry partnerships to empower digital transformation organically and providing postgraduate programmes through a unique multi-academic pan European approach. This paper identifies the needs and gaps in advanced digital skills in countries of the European Union to enable digital transformation. Validated via desk research and surveys, our results indicate that the top two advanced digital skills for companies are Cybersecurity and Cloud Computing, followed by Data Science for small and medium-sized enterprises and Artificial Intelligence (AI) for large companies, respectively. Businesses also indicate notable concerns about the digital divide, digital literacy, the potential impact of automation and job displacement, the role of regulatory frameworks, and training. Our recommendations encompass flexible and accessible education models, ex-professo AI & Data Science modules, stakeholder collaboration, and diversity & inclusion initiatives
Work Re-Entry Following Maternity Leave for First-Time Mothers: An Events, Social Identity and Intersectional Theories Informed Identity Work Framework
No full textMany first-time mothers experience significant identity issues on work re-entry following maternity leave, an important individual and life-related event. Work re-entry prompts significant identity tensions leading to identity work challenges and potential career changes. We address this significant life event and develop a subjective identity informed conceptual framework explaining its key components and outcomes. We propose that for first-time mothers, re-entry following maternity leave triggers a cognitive and subjective assessment of identity threat and opportunity leading to the use of multiple identity work strategies to address personal, role, and collective identities. We analyze the impacts of these reworked identities and identity work for career decision making and outcomes. We theoretically underpin our framework using event systems, a subjective perspective on social identity and intersectional theories and in doing so, propose future research questions and highlight implications for national policy and organizational practices
A powerful potion for a potent problem: transformative justice for generative AI in healthcare
No full textGenerative Artificial Intelligence (AI), as a transformative technology, holds significant promise for applications in healthcare. At the same time, the datafication, AI integration, and commodification of health have opened the floodgates for ethical issues, including those related to fairness, access, beneficence, democracy, solidarity, inclusion, and societal harms. As further the digitalization, innovation, and disruption of healthcare is inevitable, the paper maps out how power, equity, access, identity, participation, and knowledge contribute to creating social injustice issues. It also discusses that current justice approaches—distributive justice, representational justice, restorative justice, and capabilities-centered justice—do not have enough impact to prevent or remedy the many harms and injustices that AI has already created in healthcare or will continue to do so. The paper proposes that a transformative justice approach is needed for generative AI as a transformative technology, focused on (1) peace, emancipation, and eliminating the root causes of injustice, (2) holistic conflict resolution, (3) human rights-based approaches, and (4) the empowerment of agency and actors