Soybean is a major oilseed crop in India, and its market prices have exhibited significant volatility in recent years. Such price fluctuations create serious challenges for small and medium-scale farmers. Accurate price forecasting is essential to support informed decision-making by farmers and agri-business stakeholders. This study focuses on forecasting soybean prices in the Indian market using data spanning January 2015 to June 2025. The dataset integrates multiple heterogeneous sources, including daily market data from Agmarknet, weather information from the India Meteorological Department, regional production, and trade statistics. Extensive exploratory data analysis is conducted to examine price distributions, temporal trends, regional variations, and inter-variable relationships. Several predictive approaches are evaluated, including traditional time-series models (Autoregressive Integrated Moving Average), ensemble machine learning models, and deep learning models. In addition, a hybrid Long Short-Term Memory–Gated Recurrent Unit framework (AgroNET) is proposed to effectively model complex temporal dependencies across heterogeneous data sources. Model performance is assessed using k-fold cross-validation and evaluated through root mean squared error, mean squared error, MAPE, and R² metrics in a Python-based implementation. The results demonstrate that deep learning-based models outperform conventional approaches, with AgroNET achieving the highest R² and lowest error values. To enhance model transparency, explainable artificial intelligence using the local interpretable model-agnostic explanation technique is incorporated to identify key factors influencing individual price predictions. Overall, the proposed framework offers an effective and interpretable solution for soybean price forecasting in India and supports future research on multi-source integration and real-time agricultural price prediction. Received: 9 September 2025 | Revised: 12 January 2026 | Accepted: 11 March 2026 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement Data are available from the corresponding author upon reasonable request. Author Contribution Statement Vilas Damodhar Ghonge: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization. Yogesh Kulkarni: Conceptualization, Methodology, Investigation, Resources, Supervision.

Tomatoes are one of the most widely grown horticultural crops but are highly susceptible to leaf diseases that could significantly affect yield and quality. Early and accurate disease detection is necessary to enhance crop productivity and food security. The present study presents a deep learning system for automatic tomato leaf disease classification using a weighted ensemble of EfficientNetV2 models. Seven variants of EfficientNetV2 (B0-B3, S, M, and L) were fine-tuned on an augmented version of the filtered PlantVillage dataset. To improve feature visibility under varying lighting conditions, images were preprocessed using contrast-limited adaptive histogram equalization. Five optimization algorithms, which include Adam, Adamax, AdamW, Nadam, and RMSProp, were tested to assess their impact on model convergence and generalization. Among individual models, Adam-optimized EfficientNetV2L performed the best in accuracy with a measure of 99.50%. For classifiability resilience improvement, 16 ensemble configurations (eight unweighted and eight weighted ensembles using different combinations of EfficientNetV2-S, M, and L variants and five optimizers) were explored, with the best weights discovered using brute-force grid search. The best-performing weighted ensemble achieved 99.89% in accuracy, precision, recall, and F1-score, which was evaluated on a held-out test set not used during training or validation, demonstrating its strong generalizability. The proposed framework offers a scalable and reliable solution for early tomato disease detection with significant potential for real-time plant health monitoring in precision agriculture. Received: 10 September 2025 | Revised: 22 December 2025 | Accepted: 6 February 2026 Conflicts of InterestThe authors declare that they have no conflicts of interest to this work. Data Availability Statement Data are available from the corresponding author upon reasonable request. Author Contribution Statement Aliyu Tetengi Ibrahim: Conceptualization, Methodology, Formal analysis, Writing – original draft. Ibrahim Hayatu Hassan: Conceptualization, Methodology, Formal analysis, Resources, Data curation, Writing – original draft. Mohammed Abdullahi: Validation, Formal analysis, Resources, Data curation, Writing – review & editing, Visualization. Abeer Rashad Mirdad: Conceptualization, Software, Validation, Investigation, Writing – original draft, Writing – review & editing, Supervision, Project administration. Muhammad I. Khan: Methodology, Validation, Investigation, Resources, Data curation, Writing – review & editing, Visualization, Project administration. Saeed Ali Bahaj: Conceptualization, Software, Writing – original draft. Fatima Khan Nayer: Investigation, Visualization, Supervision, Project administration.

The transition from product-centric to service-driven paradigms in cloud computing has enabled the on-demand provisioning of scalable resources. However, the extensive operational complexity and interdependent infrastructure of cloud environments make them highly susceptible to failures, posing significant challenges to ensuring robust fault tolerance in dynamic and distributed systems. This study introduces a novel Hybrid Value-Policy Double Deep Q-Learning (DDQL) Scheduling approach, integrating value-based DDQL with policy-based reinforcement learning techniques, which are Proximal Policy Optimization and Soft Actor-Critic, to enhance fault-tolerant scheduling dynamically. Combining the advantages of both learning paradigms, the proposed method improves decision-making efficiency, enhances system resilience against failures, and optimizes scheduling performance in real-time cloud environments. Improved mean time between failures, mean time to recovery, failure recovery time, and fault tolerance success rate are guaranteed by the Hybrid DDQL scheduler's dynamic adjustments to scheduling and resource allocation algorithms. Also, it reduces the ratio of task deadline miss, energy consumption, and response time and maximizes the throughput. These results confirm the flexibility and strength of the model in dynamic and unreliable real-time cloud computing systems. Received: 27 May 2025 | Revised: 30 September 2025 | Accepted: 16 October 2025 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement Data are available from the corresponding author upon reasonable request. Author Contribution Statement Chetankumar Kalaskar: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization, Supervision, Project administration. Thangam Somasundaram: Methodology, Validation, Formal analysis, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization, Supervision, Project administration, Funding acquisition.

Online reviews play a crucial role in shaping consumer decisions, especially in the context of e-commerce. However, the quality and reliability of these reviews can vary significantly. Some reviews contain misleading or unhelpful information, such as advertisements, fake content, or irrelevant details. These issues pose significant challenges for recommendation systems, which rely on user-generated reviews to provide personalized suggestions. This article introduces a recommendation system based on a Passer Learning Optimization-enhanced Bidirectional Long Short-Term Memory network classifier applicable to e-commerce recommendation systems with improved accuracy and efficiency compared to state-of-the-art models. More specifically, the proposed model achieves a high accuracy of 98.03%, F1 score of 98.03%, precision of 98.49%, recall of 97.57%, and minimum mean square error of 1.97 based on training percentage using the patio lawn garden dataset. These results, made possible by advanced graph embedding for effective knowledge extraction and fine-tuning of classifier parameters by the hybrid PLO algorithm, establish the suitability of the proposed PLO-Bi-LSTM model in various e-commerce environments. Received: 9 April 2025 | Revised: 26 June 2025 | Accepted: 24 August 2025 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement The Amazon product data that support the findings of this study are openly available at https://cseweb.ucsd.edu/~jmcauley/datasets/amazon/links.html. Author Contribution Statement Hemn Barzan Abdalla: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Data curation, Writing – original draft, Writing – review & editing, Visualization, Supervision, Project administration, Funding acquisition. Mehdi Gheisari: Conceptualization, Software, Data curation. Awder Ahmed: Conceptualization, Methodology, Software, Formal analysis, Data curation, Writing – original draft, Writing – review & editing, Visualization. Bahtiyar Mehmed: Formal analysis, Investigation, Resources. Maryam Cheraghy: Validation, Resources, Writing – review & editing. Yang Liu: Validation, Investigation, Resources.

Deep learning techniques are transforming stock market forecasting by significantly improving the accuracy of predicting price movements and market patterns. In this research work, we propose two novel hybrid architectures – MSLSTM (Multivariate Sequential Long Short-Term Memory) and MSLSTMA (Multivariate Sequential Long Short-Term Memory Autoencoder). Both models leverage the Long Short-Term Memory (LSTM) ability to capture complex temporal dependencies in sequential financial data. Our method performs well in a variety of industries and surpasses a few other LSTMs and their autoencoder-augmented variations, as well as conventional deep learning models like CNNLSTM. MSLSTMA achieves the highest sector-wise winning rates, notably 95.83% in the telecommunication sector, 87.5% in technology, and 83.33% in industrials. Model performance was rigorously evaluated using mean squared error, root mean squared error, mean absolute error, and mean absolute percentage error. Across all these metrics, MSLSTMA consistently delivered the lowest error rates, showcasing its superior accuracy and practical relevance for real-world financial forecasting. MSLSTM also demonstrated strong performance, with winning rates of 50% in consumer staples and 45.83% in the healthcare sector. This research introduces an effective and scalable tool for investors, with the potential to enhance investment decisions through precise stock price prediction. Received: 18 March 2025 | Revised: 17 June 2025 | Accepted: 11 July 2025 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement Data sharing is not applicable to this article as no new data were created or analyzed in this study. Author Contribution Statement Sheba Elizabeth Thomas: Methodology, Validation, Investigation, Data curation, Writing – review & editing, Visualization, Supervision. Rubell Marion Lincy George: Conceptualization, Validation, Resources, Writing – original draft, Project administration. Nevin Selby: Conceptualization, Software, Formal analysis, Writing – original draft, Project administration. Aditya Taparia: Software, Formal analysis, Supervision. Jobish Vallikavungal: Methodology, Validation, Investigation, Resources, Data curation, Writing – review & editing, Visualization, Supervision.

Malware forensic investigation plays a critical role in cybersecurity, aiming to unveil malicious activities, decipher their tactics, and bolster defense mechanisms. This article introduces an innovative approach to malware forensic investigation, harnessing the capabilities of the Gray Wolf Optimization (GWO) algorithm in conjunction with a range of machine learning classifiers. These classifiers include naive Bayes, random forests (RF), decision trees, support vector machines, and K-nearest neighbors. The study leverages the CIC-MalMem-2022 dataset, which comprises memory-based data, and employs the Python programming language for model development. Research findings highlight the superiority of the RF classifier, achieving an impressive 75.6% accuracy in a multiclass classification scenario involving 16 classes. Notably, our proposed approach consistently exhibits higher accuracy when compared to existing models applied to different datasets, reaching 99.2% in binary classification. Furthermore, on the same dataset, our model outperforms the competition by achieving 86.34% and 75.64% accuracy in multiclass classification scenarios involving four classes and 16 classes, respectively. These results underscore the promising potential of our proposed model in the domain of malware forensic investigation, particularly when analyzing data extracted from memory. By combining the strength of the GWO algorithm with RF, this study aids in the progression of robust and accurate malware forensic investigation methods, thereby enhancing cybersecurity efforts in an ever-evolving threat landscape. Received: 20 December 2024 | Revised: 10 March 2025 | Accepted: 3 April 2025 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement Data are available on request from the corresponding author upon reasonable request. Author Contribution Statement Mosleh Mohammd Abualhaj: Conceptualization, Methodology, Investigation, Writing – original draft, Writing – review & editing, Supervision. Sumaya Al-Khatib: Conceptualization, Software, Validation, Resources, Data curation, Writing – original draft, Visualization. Nida Al Shafi: Software, Validation, Resources, Data curation, Writing – review & editing. Iyas Qaddara: Methodology, Formal analysis, Writing – original draft, Visualization. Abdallah Hyassat: Methodology, Formal analysis.

Increasing vehicle manufacturers has led to a considerable rise in the automotive industry's competitiveness. Various research projects are being developed to devise process optimization strategies using virtual simulation. The paper discusses the importance of applying modeling and simulation strategies to optimize automotive manufacturing processes to meet the demands of a constantly evolving market. The aim is to demonstrate a strategy divided into a few stages to optimize the process of validating new projects applied in the automotive industry. The methodology employed was computer modeling and simulation, which enabled identifying and eliminating risks inherent in the validation process, thereby reducing the time required to complete new projects. The results demonstrated that implementing this strategy facilitated an efficient transition of the validation process, maintaining the quality and safety standards required in the cell. Consequently, it can be posited that the methodology developed is an efficacious instrument for expediting the implementation of novel processes in automotive manufacturing, thereby contributing to the competitiveness and innovation of the industry in the automotive segment. Received: 24 October 2024 | Revised: 10 December 2024 | Accepted: 9 January 2025 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement Data available on request from the corresponding author upon reasonable request. Author Contribution Statement William Aparecido Celestino Lopes: Conceptualization, Methodology, Validation, Investigation, Data curation, Writing – original draft, Writing – review & editing, Visualization. Adilson Cunha Rusteiko: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization. Cleiton Rodrigues Mendes: Conceptualization, Methodology, Software, Investigation, Resources, Writing – original draft, Writing – review & editing, Visualization. Nicolas Vinicius Cruz Honório: Conceptualization, Formal analysis, Investigation, Writing – original draft, Writing – review & editing, Visualization. Marcelo Tsuguio Okano: Validation, Formal analysis, Data curation, Supervision, Project administration.

In modern life, the easy access to the internet and social media has become almost essential. Although online social media channels are good means of communication, their frequency of cyberbullying and harassment creates problems as well. On these sites, cyberbullying takes many forms that cause targeted victims to experience anxiety, guilt, and despair. With an eye on demographic factors and personal experiences of victimizing, this study explores the degree and impressions of cyberbullying within a Thai sample community. By means of a thorough questionnaire survey, important new perspectives on respondents’ awareness of cyberbullying, its underlying causes, and coping mechanisms for such events were obtained. Analyzing demographic data in tandem with cyberbullying experiences found a clear correlation between social media use length and the likelihood of running into cyberbullying. Moreover, the work suggests the creation of a customized natural language processing system specifically designed to identify Thai language cyberbullying events, therefore attaining an interesting accuracy rate of 84.23%. The study also points out the possible market chances for commercializing this method, implying it may be a profitable investment. These results provide a fresh understanding of language processing strategies for efficient content analysis on several internet platforms. Received: 16 September 2024 |Revised: 12 November 2024 | Accepted: 20 December 2024 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement Data available on request from the corresponding author upon reasonable request. Author Contribution Statement Bello Musa Yakubu: Conceptualization, Validation, Formal analysis, Investigation, Writing – original draft, Writing – review & editing, Visualization, Funding acquisition. Worapong Bumrungsri: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Data curation, Writing – original draft, Visualization. Pattarasinee Bhattarakosol: Conceptualization, Methodology, Formal analysis, Resources, Data curation, Writing – review & editing, Supervision, Project administration, Funding acquisition.

Numerous cyber threats may succeed in cloud platforms owing to unproductive authentication methods. Multiple credential authentication is a vital precautionary measure that helps reinforce cloud security for warding off imminent data breaches and illegal access. This kind of authentication approach strongly guarantees that trustworthy clients are only ratified to get cloud services, making it less tiresome for clients and more secure for organizations. This study proposes an Image Forensics-based Dual Credential Authentication (IF-DCA) based on username (i.e., preferably email ID to avoid replica) and client photographs as password. Along with a username, a rapid and acquainted user action of capturing a photograph using a web camera is adequate for this method; hence it does not need any expensive, special hardware devices. During the registration phase, along with the username, the user's face image is stored on the cloud server as a password. When registered clients need to access cloud resources or data, they should log in on the server with their username and their photograph. To improve the authentication approach, this work proposes an Attribute-controlled Conditional Generative Adversarial Network (ACC-GAN) to generate face images of the same user at various age groups. ACC-GAN includes an additional attribute control unit and an age prediction unit to synthesize photo-realistic face images with aging effects. After matching the username and face image with the registered username and ACC-GAN-generated images, the face identification module provides access to the authorized cloud user. The performance of IF-DCA is assessed through the Cross-Age Celebrity Dataset using the MATLAB R2018b/deep learning toolbox. The empirical analysis reveals that the ACC-GAN achieves better performance measures such as 98.40% accuracy, 99.3% sensitivity, 90.0% specificity, and 95.7% precision. The performance of IF-DCA is analyzed using the verification time for username, user image, and complete authentication against the number of users. Received: 4 September 2024 | Revised: 14 October 2024 | Accepted: 7 November 2024 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement The Cross-Age Celebrity Datasets that support the findings of this study are openly available at https://paperswithcode.com/dataset/cacd. Author Contribution Statement Pranali Dahiwal: Conceptualization, Software, Formal analysis, Resources, Writing – original draft, Visualization, Project administration. Vijay Khare: Methodology, Validation, Investigation, Data curation, Writing – review & editing, Supervision.

Due to increased inspection of internet traffic and the pervasive surveillance practices put in place by many organizations, virtual private networks (VPNs) are now frequently used to safeguard online privacy and circumvent censorship. However, existing VPN protocols are susceptible to methods of detection and blockage employed by network administrators, internet service providers, and in certain cases, governments. Stealth protocols and proxying techniques have therefore been presented as potential means of hiding VPN activity and avoiding discovery. Therefore, the purpose of this work is to assess how well stealth protocols and proxying work to hide the use of VPNs using a thorough experimental setup (i.e., evaluate how well-known stealth VPN protocols like Obfsproxy, Shadowsocks, and WireGuard perform against different detection techniques like deep packet inspection (DPI)). Furthermore, examine the effects of using proxy servers in conjunction with VPN connections in order to further obfuscate traffic signatures and improve privacy. The results center on how well various stealth protocols and proxying techniques defy detection attempts, offering insightful information to VPN users and developers. Thus, the outcomes help strengthen censorship resistance and online particularity in keeping pace with the development of monitoring technology. Also, analyze the efficiency of Obfsproxy, Shadowsocks, and WireGuard to provide comprehensive comparisons of current capabilities and potential enhancements. Finally, a discussion for future VPN design is conducted, focusing on the importance of adaptability in keeping secure communication channels among increasingly sophisticated surveillance measures. Received: 18 June 2024 | Revised: 2 August 2024 | Accepted: 20 August 2024 Conflicts of Interest The authors declare that they have no conflicts of interest in this work. Data Availability Statement Data available on request from the corresponding author upon reasonable request. Author Contribution Statement Mohammad Shehab: Conceptualization, Methodology, Validation, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization, Supervision, Project administration. Lial Raja Alzabin: Conceptualization, Methodology, Formal analysis, Resources, Writing – review & editing, Visualization.