Prefix sums represent a technique for precomputing partial sums to enable efficient range-sum queries. The concept is encountered under various names across disciplines: for example, summed-area tables in computer graphics, integral images in computer vision, and cumulative distribution functions in probability theory. This paper extends the use of prefix sums to triangular areas, proposing three algorithms tailored for isosceles right triangles with legs parallel to the coordinate axes. The first algorithm uses a straightforward summation of triangle rows, the second employs a recursive subdivision algorithm, while the third leverages elementary geometry principles.
In an increasingly digital world, the need for efficient and accessible content-sharing platforms has become of greatest importance. This paper introduces Share Factory, a platform that eliminates the complexities of traditional content-sharing methods by offering an array of features accessible without registration or payment. With its user-friendly interface and efficient technology stack, Share Factory platform allows users to generate custom short URLs, barcodes, and QR codes and convert colors from RGB to hex effortlessly. As the landscape of content sharing continues to evolve, such a solution plays a crucial role in shaping a more accessible, efficient, and user-centric sharing ecosystem.
AI has contributed in changing many industries, providing new and inventive solutions to complicated challenges. Nevertheless, efficient application of AI projects needs a structured and combined technique to be updated with the latest advances in the sector. There are two methodologies, CRISP-DM and OSEMN, which are used to explain the data science project life cycle on a high level. The six-phase method framework known as the Cross Industry Standard Process for Data Mining (CRISP-DM) accurately depicts the data science life cycle. On the other hand, the overall workflow performed by data scientists is categorized under the OSEMN (Obtain, Scrub, Explore, Model, iNterpret) methodology. In our study, we examine both CRISP-DM and OSEMN frameworks, and we perform a comparative analysis. We have conducted an empirical study where the experiment was organized into three case studies, each provided insightful results whether which methodology has better model fit and which has a more accurate prediction rate. The case studies suggested that CRISP-DM offers a better performance and accurate approach. All things considered, this research advances our knowledge of best methods for the selection and use of data mining methodologies, providing practitioners and researchers with direction on which strategy is best suited for their data analysis assignments.
In this contribution, we propose a method for analyzing the similarities between books considering the emotions present in their content and in online reviews, focusing on a very specific category of literature books - the novels and short stories written by Agatha Christie. The method was experimentally validated using our own reviews dataset that we collected from Goodreads and Amazon websites using our customized web scrapers. We created an experimental setup to process the book content and the book reviews towards emotion extraction and create an affective categorization of books. Lastly we discuss our research findings regarding the identified similarities between emotions conveyed by the author’s writing and those revealed in book reviews.
Detecting fraudulent activity in credit card transactions poses a serious challenge for financial institutions, which requires robust techniques that can accurately pinpoint fraudulent occurrences while minimizing false positives. In this study, we introduce an innovative strategy to enhance Credit Card Fraud Detection (CCFD) by utilizing Knowledge Graphs and Centrality measures. We propose creating a Knowledge Graph (KG) representing the credit card transaction network so as to capture connections and correlations between the transactions, and analyzing the KG to evaluate centrality measures that capture the importance of nodes and relationships within the graph. These centrality measures are utilized to enhance the input features that are used to train Machine Learning classifiers for fraud detection. Our experiments show that using the enhanced features significantly improved classification performance, providing better identification of fraudulent transactions, especially through the combination of HITS and degree centrality.
This study leverages machine learning (ML) and explainable artificial intelligence (XAI) to predict complex phenomena, specifically focusing on forecasting food security and nutritional status in Madagascar up to 2030. By combining Support Vector Machines (SVM) with SHapley Additive exPlanations (SHAP), we aim to provide both accurate predictions and transparent, interpretable insights into the decision-making process . The predictive targets include food insecurity, underweight, chronic malnutrition, and acute malnutrition, addressing critical issues related to public health and nutrition. The robustness of SVM in handling high-dimensional data, coupled with SHAP's capability to explain individual predictions, ensures that our model not only delivers reliable forecasts but also offers clarity on the importance of each feature. The methodology involves collecting diverse datasets from reputable sources, performing exploratory data analysis, and implementing SVM for predictive modeling. SHAP is then utilized to enhance model interpretability by providing detailed explanations of feature contributions. Our contributions include methodological advancements in integrating XAI with ML, development of transparent predictive models for decision-makers, and practical applications to real-world challenges in Madagascar. This research aims to support the Sustainable Development Goals by offering actionable insights for improving food security and nutrition, while also advancing global understanding of complex predictive phenomena.
This study explores the use of generative artificial intelligence (AI) in higher education from a student perspective. Using survey and interview data from business undergraduates on their main uses of generative AI, and their perceived benefits and drawbacks, this study explores AI use for the augmentation and automation of learning. Though students report AI uses that have the potential for both the automation of existing learning tasks, and for higher level learning and creativity augmentation, the main perceived benefits are focused on automation and immediate productivity outcomes. We conclude that the interplay between using generative AI for automation, augmentation or both is a useful lens for providing valuable insights for the responsible integration of generative AI in higher education.
In the digital era, the increasing reliance on web-based resources has introduced a critical yet often overlooked challenge: link rot, where hyperlinks deteriorate over time, posing a threat to the credibility and reliability of online information. To combat this challenge, we present Time Traveller, a user-friendly web tool that allows users to explore historical web snapshots. Drawing inspiration from the Wayback Machine, Time Traveller enables users to easily track the chronological changes of web pages. Through the capture and storage of web content snapshots, Time Traveller offers users the ability to revisit specific moments in internet history, ensuring the preservation of digital data. With its user-friendly interface and efficient retrieval methods, Time Traveller enhances the user experience, offering a seamless journey through the evolution of the internet, accessible to users of all technical backgrounds.
Attacks have always been conducted upon systems to bypass security layers and access sensitive data. Attacks can be intentional - performed with the explicit intention and awareness - or unintentional - performed as a result of improper data manipulation. In steganography, an important unintentional attack is represented by compression. Within this paper, we analyze steganalysis methods and image compression algorithms alongside their means of affecting input images to study challenges and dilemmas for image steganography in context of data compression. This study represents yet another fundamental step to develop a feasible and robust steganographic scheme with application in insecure network communication.
This paper presents a multi-agent system MAS-PatientCare, specifically designed to manage patient scheduling, resource allocation and diagnostic processes in hospitals. The system architecture integ-rates specialized agents, each with different responsibilities, including patient management, hospital resource allocation, scheduling, medical specialization, data preprocessing, machine learning, and decision support, to improve operational efficiency and quality of patient care across multiple hospital departments, including colonoscopy and emergency, with the flexibility to add more departments as needed. We explore the interconnectivity and collaboration between these agents, detailing how they interact to ensure seamless operations. To evaluate the effectiveness and practical applicability of the MAS-PatientCare system, it has been tested in two different situations: in emergency situations in hospitals and in the screening process for patients undergoing a colonoscopy. The impact of the MAS-PatientCare system on patient outcomes, particularly in terms of reducing waiting times, improving diagnostic accuracy and improving resource utilisation is analysed.
The development of generative AI technology has brought new prospects in several industries, including healthcare. This paper presents a case study assessing how well Microsoft Copilot, Google Gemini, OpenAI ChatGPT, and Anthropic Claude, four most popular AI systems, help with cardiology diagnosis. Analyzing Holter monitor imaging data from six patient cases, we explore the accuracy and practicality of these AI models. Our approach emphasizes the need of crafting prompts, while following ethical principles and European privacy regulations, for the production of precise and thorough AI responses. All four AI systems are shown to perform well in the study, with Google Gemini being correct most of the time. Microsoft Copilot delivers reliable results, however Claude and GPT-4o tend to overestimate patient conditions even when they identify critical parameters like drug adherence and dosage efficacy. For a three-month patient assessment, GPT-4o showed competence. This work demonstrates how generative AI might improve diagnostic procedures and seeks to add to the conversation on AI's application in patient care.
Smart homes, an important application of the Internet of Things, use the Internet to monitor and analyze data gathered from appliances within the home automation system. These intelligent appliances enable users to oversee and manage the household’s energy consumption. The smart home system collects data on the appliances’ energy consumption, generating time series data. Using this data and a time series approach, we employ the Autoregressive Integrated Moving Average (ARIMA) model to analyze and forecast energy consumption based on data from international datasets. Specifically, we selected datasets representing energy consumption per capita in the European Union (EU-27) and Romania, chosen for its integration within the EU-27, allowing for both global and local insights. The forecast generated by the ARIMA model aims to evaluate the energy required, optimizing smart home energy consumption. Accurate forecasting is critical for developing efficient energy management systems in smart homes, enabling dynamic adjustments to household energy use and significantly reducing overall energy consumption. This research underscores the importance of precise energy consumption predictions to enhance the sustainability and efficiency of smart home energy management.
Semiotics of culture is a valuable consideration, as it illuminates indiscernible aspects associated with social practices and processes of signification. Against this background, the goal of this paper is to delineate the cultural and social (contextual) dimensions of written cultural heritage from a semiotic point of view. Hence, an ontology defining a light vocabulary of high-level concepts is developed to describe these contextual dimensions. The proposed ontology, called Onto-Semiotics DCT (Describing the Contextualization of Text), is essentially a general theoretical conceptual model based on the semantic idea that a text penetrates into a communication circuit. This perspective makes a clear distinction between intra-textual and extra-textual semiotic context, as well as cultural micro-situation and social macro-situation. However, in this paper, we focus on the description of the situations that refer to the social and cultural dimensions of a text. Thus, context-aware information representation is applied. This description is crucial for the correct analysis and interpretation of texts, as according to a prevailing semiotic position, texts are influenced by the environment that produces them. As a result, knowing and recording their context reinforces text semantics, through the methodological transition from logo-centrism to culture-centrism, and even further to socio-centrism.
Because of the academic freedom they are offered, many students choose to skip a certain number of classes, which increases the absenteeism rate and affects the quality of their learning outcome and training. The goal of this paper is to propose a tool for monitoring students’ attendance and fostering their engagement and motivation for educational activities. The platform, called AttendanceManager, has three main components. The first part is a mechanism that facilitates and improves the registration of student attendance. The second component is based on gamification; taking into account the positive effects highlighted by research on this technique, the system incorporates a wide variety of badges designed to increase the engagement rate of both parties, students and teachers. The last part of the system is the analytics dashboard; teachers can visualize various student data in suggestive graphical formats and track the effects of using the platform during the semester or over the years.
More than half of the world’s population lives in urban areas and it is expected to reach 70
Through a systematic review of scientific literature, we summarize the most practical approaches for security in the application design phase. Incorporating threat modeling and secure design principles from the outset is critical to mitigating risks. Implementing secure coding guidelines helps avoid common software vulnerabilities, which means that software development teams should receive comprehensive training on secure coding techniques and integrate these practices into their workflows. In our paper we investigate practices, existing tools and literature and extract methodology to be adopted by the teams. One direction is by using IDE plugins - those can increase awareness by providing real-time feedback. Utilizing scores from tools such as Static Application Security Testing (SAST) and Software Composition Analysis (SCA) improves quality process and gives quantitative approach in decision-making during deployment, ensuring vulnerabilities are addressed early. Another direction of the research is embedding secure coding techniques in early software development lifecycle phases which helps to maintain agility in the process without affecting the release lifecycle. The third direction of research is applying the principles of separation of environments which ensure that development, testing, and production stages are isolated, reducing cross-environment contamination risks. On fourth place, we propose using the segregation of duties to further strengthen security by dividing responsibilities to prevent unauthorized access or changes. Security testing during the QA phase should include best and worst-case scenario automation, authorization matrix tests, and Dynamic Application Security Testing (DAST) to uncover potential weaknesses. Finally, assessing infrastructure vulnerability status, whether on a server or serverless level, ensures comprehensive security coverage. It is furthermore proposed to use regular external vulnerability scanning exercises which provide an additional layer of security by identifying potential threats that may have been overlooked internally. By integrating these practices, organizations can maintain a robust security posture while preserving the agility and efficiency of their development processes.
This paper presents a microservices-based distributed application designed to unify accounts management from various social networks. This application improves performance, functionality and user experience, marking a significant advance in online social networking technology. Built on a microservices architecture, the application includes noteworthy features such as a GPT 3.5-based campaign assistant designed to help users create engaging content, machine learning modules for sentiment analysis, and text summarization for posts displayed on the dashboard.
Fitness functions quantitatively assess how well a specific architecture complies with its stated architectural goals. In test-driven development, we write tests to ensure the system’s features match the intended business results. But when we turn our attention to fitness function development, we take this idea one step further and create tests that measure how well the system adheres to design goals. During our research, we found that two essential routes - applying architectural rule tests and enforcing design rules - are necessary to reach our objective. These two components cooperate to guarantee that the system’s architecture is solid and consistent with its intended purpose. We deliberately use a public domain project built with.NET to achieve this goal. This project best exemplifies a design philosophy emphasizing the division of concerns and the autonomous deployment of system components, known as Clean Architecture. This study aims to improve software engineering by encouraging the use of fitness functions earlier in the development lifecycle and improving strategic planning for cloud migrations. Creating a clear and established methodology for migration evaluation will enable this aspect. After reviewing the initial step, we introduced an additional public repository, which included creating 26 fitness functions to assess the Clean Architecture. This strategic integration allowed us to use Fitness Functions Driven Development to streamline the code reworking process, providing a smooth transition to a cloud environment. These utilities help to confirm that the product is ready for any cloud migration so that it may move to a new environment without losing any of its features or performance.
This paper presents the development and implementation of a secure, web-based decentralised marketplace for trading intellectual property (IP). By leveraging blockchain technology, this marketplace eliminates the need for intermediaries, fostering a transparent, cost-effective, and unbiased trading environment. The proposed system ensures transaction integrity and confidentiality of the traded artefacts. The key contributions include the design of a robust security protocol, the implementation of smart contracts, and the application of multiple encryption schemes to enhance security. The research underscores the potential of blockchain to revolutionise the trading of intangible creations by providing a secure and efficient platform.
Cloud computing is a prevalent technology in the IT market today, providing enterprises with the necessary infrastructure to achieve high performance levels for running their applications using the pay-as-you-go model. The lowest layer in Cloud is Infrastructure-as-a-Service (IaaS), which provides the resource pool. A primary challenge to cloud providers in IaaS is to effectively manage resources to reduce power consumption, which is a critical issue of paramount importance today, due to its substantial environmental impact stemming from the widespread use of cloud computing, while simultaneously maintaining Quality of Service. In this paper, we review the latest solutions that utilize Machine Learning (ML) algorithms for resource management. We focus on the processes of auto-scaling, Virtual Machine (VM) consolidation, and VM placement, which directly influence power consumption. We identify the benefits and innovations these solutions bring to the optimization of resource management, but also their drawbacks.
