In many countries where agriculture is a primary livelihood, farming is central to ensuring food availability and supporting economic growth. Traditionally, decisions about crop selection have been based largely on farmers' intuition and past experience. Such approaches often overlook critical parameters, including soil fertility, climatic conditions, and rainfall patterns, which strongly influence agricultural productivity. The present research presents a machine learning based crop recommendation system with the use of algorithms like Random Forest, XGBoost, and SVM (Support Vector Machine) with the goal to enhance this procedure. These models aid in the study of huge agricultural data sets and recommend the most advantageous crops for specific conditions. These mathematical models significantly lower incorrect classification rates and boost prediction efficiency when paired with successful pre processing and feature extraction methods. With a small amount of variations, the algorithm was able to accurately forecast every class in relation to its actual group distributions. With an accuracy of 99.32%, Random Forest outperformed each of the other two, followed by XGBoost at 98.60% and SVM at 96.82%. Having better decisions about what to grow is made easier by this research, This leads to more environmentally friendly farming, better utilization of resources, and greater crop yields. It displays how technology can improve agriculture's abilities, output, and capacity for future challenges.

The human eye is a vital organ responsible for vision, and the health of the retina is crucial for preserving sight. Retinal vessel segmentation plays a key role in the early detection of ophthalmic diseases such as diabetic retinopathy, glaucoma, and hypertension, where vascular abnormalities indicate disease progression. This study introduces a novel hybrid framework that enhances vessel segmentation performance using Fuzzy-Optimized UNet++ architecture, demonstrating a 2% improvement over the baseline UNet++ (95.3% accuracy). The proposed method is trained and validated on five benchmark datasets DRIVE, HRF, IOSTAR, ARIA, and CHASE_DB1, offering diversity in image resolution, pathology, and vessel morphology to evaluate cross-dataset generalization potential. To address limitations such as class imbalance, noise sensitivity, and poor micro-vessel continuity, we incorporate fuzzy logic for enhanced boundary refinement and Harris Hawks Optimization (HHO) for robust parameter tuning and convergence acceleration. Additionally, a synthetic vessel generation module, VesselGAN, is used to expand dataset diversity, achieving an SSIM score of 0.89 while preserving anatomical accuracy. Comprehensive evaluation is performed using 10-fold cross-validation and external testing on five independent datasets RETA, IDRiD, IOSTAR (external), Kaggle, and clinical-grade images. The integrated approach achieves improved performance across key metrics, including Dice Coefficient, IoU, SSIM, and F1-score, especially under noisy, low-contrast, and ultra-thin vessel conditions. This research presents a unified, end-to-end system that advances the state-of-the-art in retinal vessel segmentation. Its superior accuracy, resilience to data variability indicates robustness across unseen clinical domains and suggest suitability for real-world deployment in ophthalmic diagnostic systems.

Mobile Edge Computing (MEC) is revolutionizing computing efficiency by shifting resource-intensive and latency-sensitive operations from mobile devices with limited resources to proximal edge servers. The implementation of real-time applications is considerably improved by this prototype since it helps in reducing latency and maximizing processing effectiveness. In order to enhance Mobile Edge Computing’s potential for task offloading, this article suggests a new framework that uses Q-learning, a model-free reinforcement learning method, to dynamically handle task offloading and resource allocation in a Mobile Edge Computing environment that fluctuates. The Q-learning framework provides state-conscious decision-making to maximize long-term rewards by learning optimal techniques for workload balance and task offloading through iterative interaction with the system. The system can work out complex trade-offs between latency and resource utilization due to its flexibility. As compared to conventional techniques, this system provides context-driven accumens by apprehending the temporal and spatial dynamics of MEC environments. The proposed method is well illustrated, and a sufficient number of simulations or iterations have been made, which makes the proposed framework perform better than traditional methods. In this, we have taken parameters such as energy consumption, task execution time, and system responsiveness for our simulation. Thereby, our approach framework makes it a climbable and effective choice for approaching MEC applications. Empirical results show that Q-Learning reduces latency and improves accuracy. With the lowest latency (85 ms), Q-Learning outperforms by dynamically optimizing task offloading in response to current circumstances like network state and resource availability.

Mobile-edge computing (MEC) has evolved as a potential parameter to increase and improve the computational capabilities of mobile devices by shifting heavy resource demanded operations to adjacent edge servers. MEC is a model of distributed computing that brings cloud computing abilities nearer to users and mobile devices at the network’s edge. MEC supports basically all applications and services with minimal delay and higher bandwidth, by exploiting computing capabilities located at the edge of the network. The architecture of MEC includes many components as well as layers that support optimal computation, storage, and networking functions. In this research we proposed a deep reinforcement learning (DRL) based approach to solve the online computation offloading problem in MEC environments. Our framework proposes a real-time dynamic computational task allocation between mobile devices and edge servers based on an adaptive mechanism considering parameters such as network settings, device constraints, and server workloads. System performance was demonstrated through thoroughly simulated scenarios and experimental results shows that the DRL-based approach effectively reduced latency while maximizing energy efficiency to enhance the user experience. These results thus exhibit the promise of advanced machine learning techniques, especially DRL for efficient adaptive computation in MEC environments.

As a major global health challenge, the progressive cognitive decline and memory loss associated with Alzheimer's disease (AD) significantly impact individuals' quality of life. Due to the lack of a definitive cure, early and precise diagnosis remains crucial for implementing effective intervention and management strategies. In this study, we introduce a pioneering approach utilizing Transformer based ResLadderNet for AD classification, achieving an exceptional accuracy of 98

Diabetes prediction is a crucial focus in medical research due to the high prevalence and severe health implications of the disease. Early and accurate prediction can significantly enhance patient outcomes through timely interventions. This study examines diabetes prediction among women of Pima Indian heritage, all aged at least 21 years. Utilizing a comprehensive dataset of 2000 patient records, the study proposes an ensemble model that maximizes predictive accuracy. The approach combines the strengths of both machine learning models to build a strong predictive model. This synergistic methodology effectively balances the trade-off between model complexity and predictive power. To address potential over-fitting, an analysis model is deployed, which proved effective in ensuring the model's generalizability to new, unseen data. It demonstrates the efficacy of using ensemble models for medical predictions and highlights the potential for broader application across different populations and chronic diseases. The proposed model achieved a high accuracy of 98.37

Everybody’s life is impacted by agriculture, and crop-related disease outbreaks are a constant cause of stress for farmers all over the world. Unfavorably, the incidence of agricultural crop diseases is on the rise, resulting in considerable loss in revenue and economy. Rice is a major crop cultivated in India, but it faces numerous diseases throughout its growth stages. Rice crop leaf disease is a widespread problem, which has a negative impact on the efficiency, output, and productivity of the rice plants. Timely recognition along with precise determination and identification of bugs and illness in the rice crop, are essential for addressing this problem. The latest advances in the field of deep learning indicate that Convolutional Neural Network (CNN) can be very beneficial in these kinds of scenarios. In this paper, we used a custom CNN and ResNet-50 DL models to classify and diagnose diverse kinds of rice crop leaf diseases. An image dataset from Kaggle, which contains 5932 images of four types of rice crop leaf diseases, has been utilized to judge the outcomes. The suggested hybrid architecture, which is a combination of custom CNN and ResNet-50 architecture, consists of convolution, pooling, and dense layers. Following any required pre-processing and augmentation, the training and testing of images is carried out by the proposed architecture, and an accuracy of 95.52

Sign Language is the language used for communication involving hearing impaired and hearing disabled people that involves the movement of hands to exchange information. But even with the existence of such language, people find it difficult to communicate using the same due to its vast diversity across different regions and geographical areas of the world. For instance, ISL (Indian Sign Language) and ASL are the respective sign languages used in USA and India but they are completely different from one another from the perspective of hand signs as well as understanding. This arises the requirement for a model which provides people a basis to translate and understand ISL.The model that has been used in this work involves a pretrained model, MobileNetV2, which is further aided by fine-tuning and Transfer Learning techniques so that the model's components are reapplied to the new model thereby reducing time and computational resources. The Indian Sign Language (ISLRTC referred) dataset is employed using signs demonstrated on the ISLRTC website taken as images under different lighting conditions and backgrounds and is preprocessed and augmented thereby undergoing operations like Rescaling, Normalization, Standardization of pixels. It consists of 36 labeled classes(26 Alphabets + 10 digits) each containing a set of 1000 sample images that represent a certain gesture. The preprocessed dataset is then splitted into training and evaluation sets and the model is evaluated based on evaluation metrics that include metrics like accuracy, precision, recall and f1-scores. For better visualization purposes, confusion matrix along with graphs between accuracy and loss with epochs were plotted. An accuracy of 95.06%, precision, recall, f1-scores of 0.9438, 0.9411, 0.9410 respectively and training time of 40 minutes concluded that transfer learning balances the performance and computational cost of the model unlike other deep learning models.

The modern industries of today demand the classification of satellite images, and to use the information obtained from it for their advantage and growth. The extracted information also plays a crucial role in national security and the mapping of geographical locations. The conventional methods often fail to handle the complexities of this process. so, an effective method is required with high accuracy and stability. In this paper, a new methodology named RankEnsembleFS is proposed that addresses the crucial issues of stability and feature aggregation in the context of the SAT-6 dataset. RankEnsembleFS makes use of a two-step process that consists of ranking the features and then selecting the optimal feature subset from the top-ranked features. RankEnsembleFS achieved comparable accuracy results to state-of-the-art models for the SAT-6 dataset while significantly reducing the feature space. This reduction in feature space is important because it reduces computational complexity and enhances the interpretability of the model. Moreover, the proposed method demonstrated good stability in handling changes in data characteristics, which is critical for reliable performance over time and surpasses existing ML ensemble methods in terms of stability, threshold setting, and feature aggregation. In summary, this paper provides compelling evidence that this RankEnsembleFS methodology presents excellent performance and overcomes key issues in feature selection and image classification for the SAT-6 dataset.

Lung cancer is an extremely deadly cancer which is encountered by human beings in today’s world which also poses a significant challenge to public health. It stands among one of the major causes of deaths caused by cancer across the world and represents the highest incidence of new cases every year. With the arrival of modern diagnostic techniques and methods of treatment, the process of treatment is still very suboptimal, underlining the necessity for sustained research into the origin, early detection, prevention and treatment of this cancer. This paper explores two ML models: Support Vector Machine (SVM) and Random Forest (RF) which are used in developing an enhanced model to predict lung cancer. The dataset is assessed to measure the performance of each algorithm in predicting lung cancer severity. Key lifestyle factors influenced the models’ accuracy for lung cancer risk assessment. The RF model stood out with an accuracy of 96.32% in the prediction of risk of lung cancer, effectively handling diverse demographic and clinical variables while identifying key predictors. The SVM model demonstrated 98.56% accuracy, excelling at differentiating between early and advanced cancer stages. Cross-validation confirmed both models' reliability, while RF showed explainable risk factor analysis and SVM showed superior performance with complex diagnostic data. These machine learning approaches provide clinically valuable insights useful for improving early detection and assessing the severity of cases involving lung cancer.

Internet of Things (IoT) security and reliability rely on the capacity to identify distributed denial-of-service (DDoS) assaults in IoT networks. This research presents a comprehensive study on DDoS attack detection using the NSL-KDD dataset. The dataset contains a diverse set of network traffic data. This paper proposes two approaches, one utilizing Principal Component Analysis (PCA) and another without PCA, to compare their performance. Robust scaling and encoding techniques are applied as preprocessing steps. The experiment outcomes demonstrate a noteworthy improvement in the accuracy of DDoS attack detection in IoT devices by integrating PCA and Robust Scaler. Notably, the Random Forest and KNN classifiers demonstrate exceptional performance with an accuracy of 99.87 % and 99.14 %, respectively, while Naïve Bayes shows a lower accuracy of 87.14 %. The findings from this experiment contribute valuable insights into enhancing the security of IoT devices against DDoS attacks. The proposed approach showcases the importance of appropriate preprocessing techniques in achieving robust intrusion detection systems for IoT environments.

Digestive disorders have become very rampant these days and some of these disorders which include gastritis may cause rise in stomach polyps. It is therefore important to identify these precancerous polyps early in the disease process so that treatment can be taken to ensure they do not grow into cancer. Specifically, in the context of semantic segmentation, there has been growing concern on using deep learning methods. Among them, several segmentation models that have been introduced here have been used in various medical image segmentation tasks. With respect to our research, we investigate semantic segmentation using the U-Net3+ architecture on Kvasir-Sessile, containing 1,200 high resolution endoscopic images of sessile polyps; a diverse set of polyps that challenge segmentation due to their subtle and diverse appearance. We applied U-Net3+ to perform the segmentation of sessile polyps in the Kvasir-Sessile dataset. The increase in the ROC AUC to 0.987, MCC of 0.998, and F1-score to 0.977 from the experiments indicate that our proposed method reaches higher accurate in segmenting sessile polyps from complex endoscopic background. In addition, through qualitative analysis of the segmented images, histogram intersection, Hausdorff distance, and MSE values supported by the visual inspection confirmed precise and accurate segmentation of the various shapes and sizes of polyps.

A fractal is a pattern that emerges from the repetition of natural rules at various sizes. Fractal dimension, typically referred to as the space-filling property, offers a basis for evaluating surface roughness. The significance of fractal dimension plays a vital role in evaluating an image’s complexity, irregularity, and harshness in relation to the set to which it belongs. It is a crucial component of fractal geometry with substantial applications in many areas, such as image processing. It is used to analyse self-similar sets, such as textures and fractals. Various computational techniques exist for calculating fractal dimension, including the yardstick method, box counting method, variation method, structure function method, root mean square method, R/S analysis method, etc. However, all the commonly used methods compute the fractal dimension for binary or one-dimensional signals with the addition of grayscale images. The objective of this paper is to provide a technique for computing the fractal dimensions of colour images using a probabilistic approach without losing their colour information. We have proposed an ANN-based methodology to compute fractal dimension, and we have used tensor flow for classification of images in this application. For training, we decided to use deep learning methods because of their best pattern recognition abilities. We have also used the fractal dimension mentioned above to show that the prediction and the computed fractal dimension are interrelated.

The framework of content-based image retrieval (CBIR) is based on the visual interpretation of contents that are present in the query image, which can deal with problems like image texture classification and image texture visualization. This study uses texture features to create a CBIR system for satellite image datebase exclusively for high-resolution satellite pictures. The discussed technique makes use of a block-based scheme and the local binary pattern texture feature. In order to extract LBP histograms, the query and database pictures are separated into equal-sized blocks. The Chi-square distance is then applied to compare the block histograms. According to experimental data, the local binary pattern (LBP) formulation is a potent tool for retrieving high-resolution satellite images.

Emotion and feelings are recently becoming popular concepts in the everyday life. It not only affects human health but also plays an essential role in the decision-making processes. For this reason, emotion classification is one of the important aspects to deal with the problems like mental disorders, suicidal activities and judgmental process. Electroencephalogram (EEG) signal is one of the physiological signals which can be collected from the human brain activity while a person performing various mental and physical task. In this paper, the DEAP dataset has been implemented with the deep learning model for the classification process. In the process of developing models, for the extraction of the important features from the unprocessed EEG signals, Fast Fourier Transformation is used. Three ensemble deep learning models are tested and compared to get the best accuracy result for emotion classification. Furthermore, by the best model we can classify four emotional regions in the valance-arousal plane: HVHA, HVLA, LVHA and LVLA can be classified. The experimental results show that among all the three deep learning models, 1D-CNN-GRU achieved the highest training accuracy of 96.54