مرکز منطقه ای اطلاع رسانی علوم و فناوریJournal of Information Systems and Telecommunication (JIST) 2322-1437141202655Gated Fusion Transformer for English-Hindi Multimodal TranslationGated Fusion Transformer for English-Hindi Multimodal Translation1810.66224/jist.50948.14.1.1enPriankaSuramSchool of Computer Science and Artificial Intelligence, SR University, Warangal Telangana-506371, IndiaPramodaPatro School of Computer Science and Artificial Intelligence, SR University, Warangal Telangana-506371, India2025726Machine translation is fundamental in closing the gap between different languages, especially in the areas of con- cern and expertise such as agriculture. With the increase of digital tool usages in the agricultural practice, such an accurate and context-sensitive translation is increasingly significant. Proper delivery of agricultural information, including farm methods, weather advisories, and crop suggestions is essential among farmers, farm laborers, policymakers and researchers. Nevertheless, typical text-based translation frameworks tend to be less than optimal because of uncertainness and a restricted knowledge of context. To address these shortcomings, the proposed study refers to Multimodal Machine Translation (MMT) to incorporate textual and visual information to enhance accuracy. Gated Fusion Transformer (GFT) model has been customized to the agricultural field so that the problem of ambiguity in contexts and inconsistencies in translation can be eliminated. Training and evaluation were done using the multilingual benchmark dataset known as FLORES-200. Two commonly employed measures of performance were used, i.e. BLEU and METEOR. The system under proposal produced a BLEU of 58.2; METEOR score of 0.71, a high level and contextually relevant translation indicator. Besides benchmarking the GFT model in agricultural terms, this work adds value to the research community by offering a basis on which future development of multimodal translation systems in low-resource settings with domain-specific applications may be done.

http://jist.ir/fa/Article/Download/50948

مرکز منطقه ای اطلاع رسانی علوم و فناوریJournal of Information Systems and Telecommunication (JIST) 2322-1437141202655Transforming Public Healthcare Supply Chains: A Framework to Measure Efficiency of Heterogeneous Public Healthcare Supply Chains across Nation for Improving Drug AvailabilityTransforming Public Healthcare Supply Chains: A Framework to Measure Efficiency of Heterogeneous Public Healthcare Supply Chains across Nation for Improving Drug Availability92410.66224/jist.50942.14.1.9enAbhishekVermaAIIT, Amity University, Noida, Uttar Pradesh RekhaAgarwal AIIT, Amity University, Noida, Uttar Pradesh JitendraSinghCentre For development of Advanced Computing, Delhi2025725Health in Indian scenario is a state subject which means that states usually use their independent respective IT systems based on their specific needs and requirements. This brings a big challenge in terms of diversified nomenclatures used in across Indian states and Union Territories (UTs). Centralized data analysis is required by central agencies like Ministry of Health and Family Welfare (MoHFW) and the National Health Systems Resource Centre (NHSRC) for performance monitoring and policy making. This diversified nomenclature poses a significant hindrance in the process. The aggregation, standardization, deduplication, and visualization of data from these heterogeneous sources is both complex and resource intensive.This paper presents solution for the above challenges and propose a comprehensive framework for analyzing data from heterogeneous sources related to supply chain of drugs and vaccines. The framework incorporates fuzzy logic-based algorithms for deduplication of drug and vaccine nomenclature and supports real-time data analysis through the use of Key Performance Indicators (KPIs). It further enables centralized monitoring and decision-making via a built-in visualization layer, accessible to stakeholders at multiple administrative levels.While the framework has been tailored to the Indian public health context, its modular design makes it broadly applicable to other domains requiring integration of diverse data sources for strategic planning and policy implementation.The use of open-source technologies for development of various configurable and integrated layers like ETL, Deduplication, Standardization and Visualization layers encompassing into a single framework ensuring cost effectiveness in delivering the end-to-end solution makes it a novel and impactful for adoption specially in resource constrained environments.

http://jist.ir/fa/Article/Download/50942

مرکز منطقه ای اطلاع رسانی علوم و فناوریJournal of Information Systems and Telecommunication (JIST) 2322-1437141202655Maize Leaf Disease Detection using Deep Learning Models and a DenXNet Ensemble Model Maize Leaf Disease Detection using Deep Learning Models and a DenXNet Ensemble Model 253810.66224/jist.50993.14.1.25enMeghnaGuptaAmity Institute of Information Technology, Amity University Noida, Uttar Pradesh, India SarikaJainAmity Institute of Information Technology, Amity University Noida, Uttar Pradesh, India 000000024752339XManojKumarFaculty of Engineering & Information Sciences, University of Wollongong, Dubai, United Arab Emirates2025728Maize(Corn) is considered an important crop worldwide for global production after wheat and rice. It provides food, ethanol, carbohydrates, vitamins, and other resources, making it essential to human civilization. However, it does face numerous difficulties, including pest infestations, deteriorating soil, scarce water supplies, and climate change, resulting in various yield losses. This research introduces an efficient deep learning framework for the accurate identification of maize leaf disease. Four convolutional neural network architectures, DenXNet- MobileNet, Xception, DenseNet169, and DenseNet201- were trained and evaluated using both original and augmented datasets. To ensure fairness and eliminate data leakage, the original data is divided into train, validation, and test sets, and then augmented, whereas a stratified five-fold cross-validation strategy was applied to non-augmented data. A comprehensive ablation study was conducted to compare model performance with and without augmentation and across different ensemble configurations. The study explored soft ensemble modelling using combinations of two and four base models. Among all configurations, the four-model ensemble, DenXNet, achieved the highest accuracy of 98.46% and consistency across folds, outperforming individual and partial ensembles. The proposed method demonstrates improved precision, reduced over fitting, and strong adaptability for real-world agricultural disease detection tasks.

http://jist.ir/fa/Article/Download/50993

مرکز منطقه ای اطلاع رسانی علوم و فناوریJournal of Information Systems and Telecommunication (JIST) 2322-1437141202655Detecting Synchronized Hate Speech in Online Social Networks via Social Synchrony and Ant Colony OptimizationDetecting Synchronized Hate Speech in Online Social Networks via Social Synchrony and Ant Colony Optimization394910.66224/jist.50944.14.1.39enShabanaNargis RasoolDepartment of Computer Science, Islamic University of Science and Technology, Kashmir, India 0000-0003-2512-0204SarikaJainAmity Institute of Information Technology, Amity University Noida, Uttar Pradesh, India000000024752339XAjay VikramSingh Department of Computer Science, Islamic University of Science and Technology, Kashmir, India 0000-0003-4129-647X2025725Online platforms have become fertile grounds for hate speech, often spreading through bursts of coordinated user activity. Detecting such patterns requires more than analyzing individual posts, as it calls for understanding the collective rhythm of online interactions. In the present study, we present SIACO (Social Synchrony Identification using Ant Colony Optimization), a nature-inspired framework that detects hate-speech events by tracing synchrony in user behaviour. SIACO models how hateful expressions emerge and fade collectively, using Ant Colony Optimization to refine linguistic features and improve classification accuracy. Upon evaluation on a Twitter dataset, the framework consistently outperforms both traditional machine learning models and transformer-based baselines, achieving up to a 10% improvement across major evaluation metrics. The framework also offers interpretable insights into the linguistic and temporal cues driving coordinated hate. The performance scores obtained highlight the value of looking at hate speech not just as text, but as a social phenomenon unfolding in synchrony.

http://jist.ir/fa/Article/Download/50944

مرکز منطقه ای اطلاع رسانی علوم و فناوریJournal of Information Systems and Telecommunication (JIST) 2322-1437141202655Optimized Gradient Boosting for Financial Forecasting: A Data-Driven Approach to Gold Stock PredictionOptimized Gradient Boosting for Financial Forecasting: A Data-Driven Approach to Gold Stock Prediction505910.66224/jist.51042.14.1.50enShreyaGaragComputer Science Department, Christ University, Bengaluru, IndiaJossyGeorgeComputer Science Department, Christ University, Bengaluru, India Akhil M.NairLuxsh Technologies Pvt Ltd, United KingdomBosco PaulAlapatt Computer Science Department, Christ University, Bengaluru, IndiaRiyaBaby Computer Science Department, Christ University, Bengaluru, India2025731The application of machine learning algorithms in finance forecasting and stock investment domain has revolutionized the way the financial data is analyzed, interpreted and employed for various investment options. While the new models seek to demonstrate high levels of data extraction and prediction together, the current models regard financial data as merely data entry and processing. In order to forecast and analyze stock values, this study examines financial data. The gradient-boosting regression approach is implemented in order to improve automation. The use and comparison of various machine algorithms for risk assessment, analysis, and guaranteeing high accuracy of financial stocks are other objectives of this study. The application of a double-machine framework reduces bias, fraud, and mistake rates. Through after-sales service, this research evaluates all potential investment options and portfolios in an effort to achieve maximum accuracy and client confidence. Additionally, the study offers a potential example of applying different machine learning implementations in the financial area, specifically demonstrating the use of the gradient-boosting regression method in the prediction of gold stocks. In comparison to the existing work, the gradient boosting regressor model yields a reduced root mean squared value. The dataset was imputed using median and features with more than 30% missing values were removed for further processing The proposed work demonstrates high predictive accuracy and reduced root mean squared value support our proposed work for more dependable forecasting when it comes to stock price prediction.

http://jist.ir/fa/Article/Download/51042

مرکز منطقه ای اطلاع رسانی علوم و فناوریJournal of Information Systems and Telecommunication (JIST) 2322-1437141202655Enhancing Industrial Interaction Practices Through AI-Based Parameter Modeling Enhancing Industrial Interaction Practices Through AI-Based Parameter Modeling 606910.66224/jist.50946.14.1.60en Ashwini KumarAmity Institute of Information Technology, Amity University, Uttar Pradesh, Noida, IndiaRekhaAgarwalAmity Institute of Information Technology, Amity University, Uttar Pradesh, Noida, IndiaArchanaSinghMinistry of Education, Government of India2025725Industrial systems today depend increasingly on effective communication and coordination among humans and machines. This study proposes an Artificial Intelligence–based approach for modeling and improving industrial interaction practices using the COFI framework—Context, Content, Competency, and Culture. By combining supervised and unsupervised learning techniques, specifically Random Forest (RF) and K-Means clustering, the research models key parameters that influence communication efficiency and organizational alignment. A publicly available behavioral dataset, supplemented with simulated industrial communication records, was used to represent multi-agent interactions within a workplace context. Extensive data preprocessing, feature engineering, and COFI-based variable mapping were performed to ensure interpretability and conceptual coherence. The RF model achieved an improved predictive accuracy of 72.4% following feature optimization, while K-Means clustering produced three distinct communication groups with a Silhouette score of 0.75 and a Davies–Bouldin Index of 0.49, indicating well-separated clusters. Feature-importance and SHAP analyses revealed that contextual and content-based variables contributed most significantly to prediction outcomes, while competency and cultural attributes shaped nuanced interaction patterns. A pilot case simulation demonstrated tangible performance improvements—reducing response time by 12% and improving task resolution by 9% when AI insights were applied to industrial communication workflows. The findings confirm that combining supervised prediction with unsupervised segmentation offers a robust pathway to understanding and optimizing human–machine communication within organizational ecosystems. This research contributes a practical and interpretable framework for AI-enabled industrial interaction modeling, offering both theoretical insight and applied value for adaptive, data-driven management systems.

http://jist.ir/fa/Article/Download/50946