Journal of Data Engineering and Knowledge Discovery

A New Method of Portfolio Selection Using Mutual Information and MST

2026-02-16T12:12:54+00:00

The stock market fluctuates randomly since it is a complex system which is formed by the interaction of individual factors. One of the approaches to researching such an interaction-based complex system is network analysis. The current paper proposes a new portfolio selection method utilizing the structural information of the network. First of all, the co-relationship between assets is worked out using mutual information, and on this basis, MST is generated. Next, the scores of three centralities - degree centrality, closeness centrality, and betweenness centrality are calculated on the network. Finally, various portfolios are constructed based on the rankings of the centralities of each asset, and their effectiveness is examined. The findings of the experiment showed that the portfolios composed of stocks with low centrality scores produce higher returns, and this proved that there is a negative relationship between the centrality of assets and the weight of portfolios in the network and verified the usefulness of mutual information in the network analysis.

Enhancing News Authenticity Prediction with Machine Learning Approaches

2026-03-02T12:08:40+00:00

The issue of misinformation becoming prevalent in social media platforms has now taken a turn and has become a significant threat to public discourse, democratic processes, and societal trust. Manual verification systems simply cannot keep pace with the sheer volume and velocity of fake news being generated, thus necessitating the activation of automated detection mechanisms. This time, a technique based on machine learning for the purpose of detecting false news with the passive-aggressive classifier plus term frequency-inverse document frequency (TF-IDF) vectorization is introduced. The model developed underwent training and evaluation on the basis of a benchmark dataset of 44,898 news articles, among which 23,481 were marked as fake while 21,417 were real. The system proposed was able to obtain 99.48% accuracy, reporting precision and recall scores of 0.99 for both classes, thus proving its effectiveness in differentiating the fabricated content from the real one. The algorithm is suitable for this task due to the fact that it requires minimal computational resources, is low-weight, thus allows the model to be deployed in real-time, and is also more cost-efficient than deep learning, which is resource-demanding. The strength of the proposed approach was established through the implementation of confusion matrix analysis and standard classification metrics. The purpose of this project is to connect the theoretical research with real-life use by creating an efficient and easily usable application for the counteraction of digitally spread misinformation.

An Intelligent Virtual World Framework Using Computer Vision and Speech Control

2026-03-26T17:04:12+00:00

Traditional computer interaction systems mainly rely on keyboards, mouse, or external controllers, which restricts natural and immersive human–computer interaction. Consequently, traditional interaction methods are not well-suited for modern applications such as virtual environments, training systems, and assistive technologies. Hence, there is a need to develop an interactive system that enables users to control virtual environments through natural human actions such as gestures and voice commands. This research presents a multimodal virtual world system that integrates computer vision and speech recognition technologies to enable hands-free interaction within a digital environment. The proposed system captures real-time video input through a camera to detect hand gestures using computer vision techniques and processes voice commands through speech recognition. These inputs are mapped to predefined actions within the virtual environment, enabling smooth and intuitive control. The system is developed in Python, using OpenCV for real-time image processing and speech recognition libraries to identify and process voice commands. The integration of gesture-based and voice-based interaction improves accessibility and enhances the overall user experience. Experimental results show that the system performs efficiently in real-time with accurate gesture and voice recognition under normal conditions. The developed system is affordable, user-friendly, and applicable to domains such as education, gaming, virtual training, and human computer interaction.

An Artificial Intelligence-based Integrated Framework for Oceanographic, Fisheries, and Molecular Biodiversity Data

2026-04-06T12:07:11+00:00

Large amounts of ocean, fisheries, and biodiversity data are collected today from many sources, such as satellites, fishing records, environmental sensors, research articles, and DNA sequencing labs. Even though this information is valuable, it remains scattered in different formats and locations. Because of this breakage, it becomes difficult for fishermen, researchers, and policymakers to access the knowledge they need for sustainable marine planning and decision-making. This project introduces an AI-based unified data platform that brings all ocean-related knowledge into one system. The platform will collect, clean, and organise multi-modal data from satellite imagery, CSV files, scientific literature, and molecular datasets. Fishermen will receive forecasts of safe and high-probability fishing zones. The unified dashboard will provide user-specific outputs. Policymakers will obtain insights into overfishing risk areas and conservation requirements. The goal is to support both economic gains and ecological sustainability. By combining real-time analytics with spontaneous visual tools, this platform will create a bridge between scientific data and field-level decision-making.

Revisiting Object-oriented Design in Modern Java: An Evolutionary Perspective

2026-04-07T06:17:07+00:00

Object-oriented programming (OOP) has long been an essential approach in software development, and Java has consistently remained a widely used programming language that supports this paradigm. Over the years, Java has evolved to meet modern software requirements while continuing to follow object-oriented principles. Recent advancements in Java have introduced new language features that significantly influence object-oriented design practices. These developments affect how core OOP concepts such as encapsulation, inheritance, abstraction, and polymorphism are implemented in real-world applications. Modern Java encourages developers to adopt clearer and more structured design approaches. This study focuses on features such as records, sealed classes, and improved interface capabilities. These constructs help reduce repetitive code, support immutable data handling, and restrict improper inheritance, leading to more controlled and predictable class designs. By analyzing these enhancements, the paper highlights how modern Java promotes cleaner code structure, improves program safety, and enhances long-term maintainability. The study concludes that while the foundational principles of object-oriented programming remain unchanged, recent Java features provide developers with more efficient and well-organized tools to apply object-oriented design in contemporary software systems.