Explore the vast range of titles available.

A larger proportion of crops face disease outbreaks, making agricultural output difficult. Detecting and predicting diseases at an early stage can enhance productivity. Guava, a tropical and subtropical fruit, is cultivated in various countries. In regions such as Bangladesh, Pakistan, India, and South America, guava cultivation faces significant challenges due to diseases like Canker, Dot, Mummification, Phytophthora, Scab, and Styler and Root. Traditional diagnosis methods based on visual observation are often unreliable and time-consuming. To address this, we developed an automated system leveraging deep learning techniques. Our study utilized a dataset comprising 4046 guava leaf images categorized into these seven disease classes. We compared the performance of traditional methods with deep learning approaches using vision transformers and transfer learning. The results demonstrate the superiority of deep learning methods over traditional approaches, where traditional machine learning model SVM gave accuracy near 78% and deep learning methods gave over 90%. The transfer learning method gave an accuracy of nearly 97% and on the other hand, the vision transformer gave accuracy of 98%. This offers a promising solution for early disease detection in guava crops.

Assessment is an essential part of education, both for teachers who assess their students as well as learners who may evaluate themselves. Multiple-choice questions (MCQ) are one of the most popular types of knowledge assessment, e.g., in medical education, as they can be automatically graded and can cover a wide range of learning items. However, the creation of high-quality MCQ items is a time-consuming task. The recent advent of Large Language Models (LLM), such as Generative Pre-trained Transformer (GPT), caused a new momentum for automatic question generation solutions. Still, evaluating generated questions according to the best practices for MCQ item writing is needed to ensure docimological quality. In this article, we propose an analysis of the quality of LLM-generated MCQs. We employ zero-shot approaches in two domains, namely computer science and medicine. In the former, we make use of 3 GPT-based services to generate MCQs. In the latter, we developed a plugin for the Moodle learning management system that generates MCQs based on learning material. We compare the generated MCQs against common multiple-choice item writing guidelines. Among the major challenges, we determined that while LLMs are certainly useful in generating MCQs more efficiently, they sometimes create broad items with ambiguous keys or implausible distractors. Human oversight is also necessary to ensure instructional alignment between generated items and course contents. Finally, we propose solutions for AQG developers.

Between 1990 and 2023, Chile’s Congress processed and approved 2738 laws, with an average processing time of 667.8 days from proposal to official publication. Recent political circumstances have underscored the need to identify legislative proposals that can be expedited for approval and which ones are unlikely to be approved at all. This article describes a bottom-up, data-driven classification of voting (and voters) on law proposals, which yield two axis: polarization (lack of agreement on an issue), and (political) alignment (intra-party coincidence of a group’s members regarding certain opinion). And four quadrants: “ideological stance” (high polarization, high alignment), “personal interests” (high polarization, low alignment), “thematic interest” (low polarization, low alignment), and “technical consensus” (low polarization, high alignment). We used this scheme to analyze an existing Open Linked Dataset with semantic web technologies (ontologies, RDF Shape expressions, and URI patterns), which records parliamentarians’ political parties and their voting on law proposals during 1990–2023. We found that most bills (70.14%) are in the technical consensus quadrant, and could have been quickly shepherded to approval. Wider adoption of this analysis to classify new bills may help to speed up their legislative processing, ultimately allowing Congress to serve citizens in a more timely manner.

The successful development of an effective machine learning model for detecting malicious queries is a challenging task that requires domain expertise. Malicious queries are complex and constantly evolving, and their detection is further complicated by variations in database server architectures. To tackle this challenge, domain experts with a deep understanding of attack vectors, techniques, and trends are crucial. Their insights into the patterns, behaviour, and characteristics of malicious queries provide invaluable guidance for designing appropriate features and training the model. This study aims to leverage this domain expertise to propose a robust machine-learning model capable of accurately identifying and categorizing malicious queries. By incorporating expert knowledge and using advanced techniques like machine learning, the model can be optimized to address the dynamic nature of these threats, leading to improved accuracy in real-world scenarios. After the extensive experiment, using 88,213 malicious and normal queries from GitHub and other sources, the Random Forest model yielded the highest accuracy of 98.4% and 97% sensitivity compared to other machine learning models experimented in the study. The performance comparison of the study with existing works indicates that the machine learning model achieved promising results.

Smart contracts, widely utilized in blockchain applications, are vulnerable to security threats that can lead to severe financial and operational consequences. This study introduces the Hierarchical Contextual Embedding with Hybrid Deep Ensemble Network (HCE-HDEN) framework, designed to enhance the detection of critical vulnerabilities, specifically Integer Overflow, Reentrancy, and Timestamp Dependency. The proposed methodology integrates semantic learning techniques with hierarchical code tree architectures, significantly improving detection accuracy by capturing complex contextual relationships within smart contract code. To evaluate its effectiveness, HCE-HDEN was tested against nine state-of-the-art vulnerability detection tools, including SmartCheck, Manticore, Osiris, Oyente, ES, and Slither. Experimental results demonstrate that HCE-HDEN outperforms existing tools, achieving superior detection rates of 75% for Integer Overflow, 95% for Reentrancy, and 88% for Timestamp Dependency, compared to ES, the closest competitor, which achieved detection rates of 64%, 90%, and 80%, respectively. While ES exhibited the fastest execution time at 0.16 seconds, HCE-HDEN maintained a balance between accuracy and efficiency, processing smart contracts within 0.8 seconds. The study further highlights that the optimization strategies employed in HCE-HDEN, such as parallel-processing neural networks and hierarchical code structures, significantly enhance vulnerability detection while ensuring reasonable processing speed. These findings establish HCE-HDEN as a highly effective framework for smart contract security analysis, offering superior accuracy over existing automated tools. Future research will focus on further optimizing execution efficiency while expanding vulnerability detection capabilities to address a broader range of security threats in blockchain-based applications.

Cloud Computing (CC) is the model of delivering services for users across internet. The CC is useful to address various issues like scheduling, security and Load balancing (LB). Amongst these problems, LB is the most demanding issue. LB is performed at VM or PM level. When the collection of tasks enter into VM, it uses VM resources and these resources become exhausted, that indicates that there are no resources available for handling the other task requests. Here, an efficient model named Adam_Pufferfish Optimization Algorithm (Adam_POA) is developed for LB in CC. Firstly, the tasks are assigned to the VM in the Data Center (DC) in round robin manner. Based upon VM parameters, VMs are classified as overloaded and underloaded VMs employing Deep Fuzzy Clustering (DFC). After that, divide the tasks in overloaded VM based upon priority. Following to this, the tasks are assigned in overloaded to underloaded VMs for balancing the load in the cloud using hybrid Adam_POA. It is recognized that Adam_POA obtained resource availability with 0.880, capacity with 0.915, load of 0.529 and 0.857 of reliability.

There is huge untapped potential for the application of Large Language Models in the fields of marketing and advertising. In this work, we describe an approach to automate the generation of brand territory maps , a visualization used by advertising strategists to understand how a brand is perceived by consumers and differentiated from its competitors. We collect a Household Item data set with product reviews to exemplify our approach. We elicit customer perceptions from ChatGPT with regards to certain dimensions, viz. , esteem, reliability, modernity, quality and relevance. By analyzing customer reviews using this Large Language Model, we show that it is possible to get a broader view of how consumers perceive specific aspects of certain products or brands in an automated fashion. We perform an empirical evaluation to compare the brand territory maps generated using our approach and those generated by humans. We find that (1) Human responses have significantly more variation than those by ChatGPT; (2) Making small adjustments to the ChatGPT prompt can result in stark differences in results; (3) Our approach is extremely cost effective compared to that of using humans to generate brand territory maps; and that (4) The brand territory maps generated using our approach are comparable to those generated by humans.

Data encryption which is associated with cryptography is necessary to prevent the compromise of Personally Identifying. Multi-level security is ensured by combining the Huffman code with certain cryptographic techniques, such as symmetric encryption algorithms. In order to decode the message, Huffman code can access both the code wordlist and the encoded bits. Inadequate error-correcting techniques with encoded bits can be harmed in this way, leading to the complete loss of information. Nevertheless, sending code wordlists and bits of code to the recipient takes a lot of transmission time. It is necessary to offer strategies for countermeasures to secure and fix the broken encoded bit. So, the problem is to safeguard the original wordlist from tampering without compromising the current code bit. Information. The goal of the investigation is to reduce interference in the levels of communication between the reader and the card. Due to the card's susceptibility, a Mifare classic 1K and Radio Frequency Identification (RFID) were used in this study's simulation. The experiment when N  = 200 was conducted and the result below was obtained. The number of unique characters at 0.5 in AES running mode with 0.023 ms at CBC optimal and ECB 0.023 ms. It was possible to simulate the Advanced Encryption Standard (AES) using changing bytes and a number of unique characters.

Proper recognition and classification of medicinal plants is fundamental to pharmacology, botany, and traditional medical applications. However, distinguishing closely related species remains difficult, and traditional methods have not been able to fully address these challenges. In this study, we propose a pioneering deep learning architecture, the Hierarchical Contextual Vision Integration Network (HCVINet), which integrates multi-level image feature extraction with contextual-semantic information using natural language processing (NLP) techniques. HCVINet employs a Hierarchical Feature Extraction Network to capture a broad spectrum of visual features—from elementary textures to complex patterns—and fuses these with semantic data through a Contextual-Correlation Integrated Network (CCINet), enabling the model to utilize both visual and textual information for improved classification decisions. Experiments conducted on benchmark medicinal plant datasets demonstrate that HCVINet achieves a classification accuracy of up to 98.3%, outperforming existing state-of-the-art CNN-based models by an average margin of 4.5%. The model also yields higher retrieval rates and improved harmonic mean scores, validating the effectiveness of combining visual and semantic cues. While HCVINet proves to be a robust tool for medicinal plant classification, its performance may be influenced by the quality of the textual corpus, and occasional mismatches between text and image features can affect classification outcomes. Overall, HCVINet offers a significant advancement in automated plant identification and provides promising implications for research and practical applications.

Liver cancer remains a leading cause of cancer-related mortality worldwide, and timely, accurate classification of liver tumors in magnetic resonance imaging (MRI) is critical for improving patient outcomes. However, automated liver tumor classification is challenged by blurred tumor boundaries, variable lesion sizes, and heterogeneous image characteristics. To address these issues, we propose a novel Convolutional Gated Kronecker Network (CGKN) framework that integrates spatial and contextual feature learning for robust liver tumor classification in MRI. The methodology begins with preprocessing MRI images using an Adaptive Wiener Filter to enhance image quality and reduce noise, followed by precise segmentation of liver and lesion regions using a Dynamic Context Encoder Network. After extracting relevant features, the CGKN—combining Convolutional Neural Network-Gated Recurrent Unit and Deep Kronecker Network modules—classifies tumors into mild, moderate, or severe categories. This hybrid approach is designed to overcome the limitations of existing models, such as limited generalizability, high computational cost, and insufficient integration of spatial and contextual cues. Experimental evaluation on a benchmark MRI dataset demonstrates that the CGKN achieves an accuracy of 92.88%, a true positive rate of 91.64%, and a true negative rate of 91.56%, outperforming several state-of-the-art deep learning models. These findings highlight the potential of the CGKN framework as an effective and efficient solution for automated liver tumor classification in MRI, supporting more reliable computer-aided diagnosis and facilitating clinical decision-making.