Nadim Rana

Accurate and interpretable solar power forecasting is critical for effectively integrating Photo-Voltaic (PV) systems into modern energy infrastructure. This paper introduces a novel two-stage hybrid framework that couples deep learning-based time series prediction with generative Large Language Models (LLMs) to enhance forecast accuracy and model interpretability. At its core, the proposed SolarTrans model leverages a lightweight Transformer-based encoder-decoder architecture tailored for short-term DC power prediction using multivariate inverter and weather data, including irradiance, ambient and module temperatures, and temporal features. Experiments conducted on publicly available datasets from two PV plants over 34 days demonstrate strong predictive performance. The SolarTrans model achieves a Mean Absolute Error (MAE) of 0.0782 and 0.1544, Root Mean Squared Error (RMSE) of 0.1760 and 0.4424, and R² scores of 0.9692 and 0.7956 on Plant 1 and Plant 2, respectively. On the combined dataset, the model yields an MAE of 0.1105, RMSE of 0.3189, and R² of 0.8967. To address the interpretability challenge, we fine-tuned the Flan-T5 model on structured prompts derived from domain-informed templates and forecast outputs. The resulting explanation module achieves ROUGE-1, ROUGE-2, ROUGE-L, and ROUGE-Lsum scores of 0.7889, 0.7211, 0.7759, and 0.7771, respectively, along with a BLEU score of 0.6558, indicating high-fidelity generation of domain-relevant natural language explanations.

The detection of maize plant leaf diseases is a critical aspect of agricultural management, necessitating accurate and efficient methodologies. The field of maize plant leaf disease detection encounters several challenges that hinder the development of robust and effective solutions. One prominent challenge is the diversity of disease indicators, wherein various pathogens can cause similar symptoms, making it challenging to differentiate between diseases accurately. Additionally, limited annotated datasets pose a constraint, hindering the training of deep learning models and potentially leading to suboptimal generalization. The dynamic nature of plant growth and environmental conditions further complicates disease detection, as the appearance of symptoms may vary at different stages of plant development. Another challenge lies in achieving a balance between accuracy and computational efficiency, especially in real-time applications, as many existing models struggle to provide rapid and precise results simultaneously. This paper addresses prevailing challenges in disease detection by introducing a customized CenterNet approach, incorporating DenseNet-65 as its foundational architecture. Common issues encountered in traditional disease detection models are addressed through our proposed approach, which demonstrates superior performance in both disease classification and precise localization of affected regions. A comparative analysis is conducted against conventional and contemporary methodologies, emphasizing the innovation and competitiveness of our model in advancing maize plant leaf disease detection. We evaluated our approach on a standard dataset CD&S and attained a classification accuracy of 98.62%. This contribution not only expands the current understanding of disease detection in agriculture but also offers a practical and efficient solution for improved maize crop management.

Thoracic diseases, including pneumonia, tuberculosis, lung cancer, and others, pose significant health risks and require timely and accurate diagnosis to ensure proper treatment. Thus, in this research, a model for thorax disease classification using Chest X-rays is proposed by considering deep learning model. The input is pre-processed by resizing, normalizing pixel values, and applying data augmentation to address the issue of imbalanced datasets and improve model generalization. Significant features are extracted from the images using an Enhanced Auto-Encoder (EnAE) model, which combines a stacked auto-encoder architecture with an attention module to enhance feature representation and classification accuracy. To further improve feature selection, we utilize the Chaotic Whale Optimization (ChWO) Algorithm, which optimally selects the most relevant attributes from the extracted features. Finally, the disease classification is performed using the novel Improved Swin Transformer (IMSTrans) model, which is designed to efficiently process high-dimensional medical image data and achieve superior classification performance. The proposed EnAE + ChWO+IMSTrans model for thorax disease classification was evaluated using extensive Chest X-ray datasets and the Lung Disease Dataset. The proposed method demonstrates enhanced Accuracy, Precision, Recall, F-Score, MCC and MAE of 0.964, 0.977, 0.9845, 0.964, 0.9647, and 0.184 respectively indicating the reliable and efficient solution for thorax disease classification.

Due to the extensive migration of business and scientific applications as well as the enormous growth in online data produced by IoT devices, numerous problems have arisen in cloud scheduling. Efficient delivery of resources considering user-defined Service Level Agreement (SLA) and Quality of Service (QoS) can only achieve with efficient and state-of-the-art scheduling methods. In this regard, virtual machine (VM) scheduling has been a highly required method for resource scheduling in the ever-changing cloud and multi-access computing environment (MAC). Based on an examination of recent literature, this investigation intends to provide a comprehensive Systematic Literature Review (SLR) of the methods employed for virtual machine scheduling in cloud computing. Besides, the SLR disseminates the challenges and opportunities in VM design and discusses future researchers' baselines. The SLR investigated the VM scheduling techniques and searched the most relevant research databases online. The authors selected sixty-seven (67) preliminary studies for this review out of 722 articles between 2008 and 2022. A total of 67 articles were reviewed for VM scheduling methods and techniques. The taxonomical results were divided into three major classes; conventional approach, heuristics approach, and meta-heuristic approach. With the observation, this review concludes that a lot of development in VM scheduling techniques in the literature are based on metaheuristics and heuristics methods. At last, many open issues, challenges, and development trends of modern VM scheduling techniques are discussed.

The use of modern technologies for control and monitoring and accessing devices in domestic or industrial buildings with convenience, comfortable and easy access from any location is the primary aim of the internet of things (IoT) technology for smart home automation. Complete Smart home automation, with overall control from any place at any time is still not fully available. Nevertheless, this work proposes a mobile application system for smart homes, with the purpose of overall monitoring and control of home appliances and devices. The proposed method is based on Zigbee, Arduino and Bluetooth for wireless communication among devices in the home. At the same time, a mobile application is used for the control and monitoring of the devices or appliances. In this study, Zigbee and Bluetooth are combined in order to establish efficient communication either within or outside the home premises. A user scenario of the proposed work was simulated using Proteus Simulation software to validate the practicability of the new system.

More often than not, commuters are left stranded at pick-up spots – clueless about the availability and proximity of public transport vehicles hence the stigma of public transport being unreliable, especially in developing countries. This is a result of poorly managed fleets, caused by varying demands and rigid schedules. In this paper, we present an intelligent real-time transport information system to keep commuters informed about the status of buses currently in transit, and also provide an insight to bus managers based on ridership data and commuter behavior. The system is composed of three subsystems designed to cater for commuters, bus-drivers and bus managers respectively. This system is developed on the Backend-as-a-Service (BaaS) platform Firebase. Furthermore, a neural network is trained to provide predictions to bus managers on the expected ridership numbers per route. The trained model is integrated with a web application for bus managers. An Android application used by bus drivers collects the ridership data being fed to the network. The proposed system was evaluated with a real-world data set that contains the daily ridership on a per-route basis dating back to 2001. Evaluation results confirm the effectiveness of the new system in reducing the total mileage used to deliver commuters, reducing fuel costs, increasing the profit of bus operators, and increasing the percentage of satisfied ridership requests.