The increasing use of electronic health records (EHRs) has transformed healthcare management, yet data sharing across institutions remains limited due to privacy concerns. Federated learning (FL) offers a privacy-preserving solution by enabling collaborative model training without centralized data sharing. However, non-independent and identically distributed (non-IID) data distributions, where the data across clients differ in class proportions and feature characteristics, pose a major challenge to achieving robust model performance. In this study, we propose a hybrid framework that combines the Federated Proximal (FedProx) algorithm with the ResNet50 architecture to address non-IID data issues. We artificially partitioned an IID brain tumor dataset into non-IID subsets to simulate real-world conditions and applied data augmentation techniques to balance class distributions. Global model performance is monitored across 100 training rounds with varying regularization parameters in FedProx. The proposed framework achieved an accuracy of 97.71% on IID data and 87.19% in extreme non-IID scenarios, with precision, recall, and F1-scores also demonstrating strong performance. These findings highlight the effectiveness of combining data augmentation with FedProx in mitigating data imbalance in FL, thereby supporting equitable and efficient training of privacy-preserving models for healthcare applications.

Medical image classification using deep learning (DL) typically requires large and diverse datasets. However, data privacy regulations often limit data sharing across institutions. Federated Learning (FL) addresses this issue by enabling collaborative model training without transferring raw data. Despite its advantages, FL is challenged by limited data at each participating client, which can hinder model performance. To overcome this limitation, we employ Federated Transfer Learning (FTL), a hybrid approach that combines FL with Transfer Learning (TL) to improve model generalization under data scarcity. In this work, we apply FTL to chest X-ray (CXR) classification, leveraging MobileNet for one dataset and ResNet50 for another. We have evaluated our framework’s performance using various evaluation metrics. It achieved 98% accuracy and 99.97% AUC-ROC on Dataset1, and 93.46% accuracy with a 97.9% AUC-ROC on Dataset2, demonstrating its overall effectiveness. To enhance model interpretability, we use Explainable AI (XAI) techniques such as Grad-CAM and LIME to visualize decision-making. Furthermore, we employ two different GPT models-Gemini and ChatGPT-one for generating human-readable explanations based on the XAI visualizations and the other to quantitatively validate the reliability of the generated explanations on a five-point Likert scale. The proposed approach yielded reliability scores of 4.13 and 4.20 for GradCAM visualizations, and 4.43 and 4.87 for LIME visualizations, across the two datasets, indicating high reliability. Overall, the proposed FTL-XAI-GenAI framework ensures high classification performance and transparency, enabling medical professionals to understand AI-driven diagnoses while maintaining data privacy.

This article describes the architecture and system design for assisting blind people in navigating freely inside an enclosed environment, such as the home or the outdoors. Thus, the proposed technology uses IoT technology and emerging techniques for machine learning to provide high-tech cane functionality that allows visually impaired navigators to walk independently. It also includes mobile applications to safeguard visually impaired persons and allow guardians to observe them. The proposed in this study system is intended to identify and classify any obstacles within a defined distance using machine learning. In this connection, an indoor and outdoor architecture on YOLO v3 is implemented for its detection technique, and multi-layer perceptron (MLP) neural network technology supports this framework. Based on the detection and classification, YOLO v3 and MLP are crucial for their accuracy.

Topic analysis is also known as topic detection or topic extraction, refers to ML method that categorizes larger text datasets into categories based on the individual text. It employs natural language processing to analyze human communication by breaking it down into components such as speech, words, sentences, and context, aiming to identify patterns and unveil underlying meanings within texts. This process aids in deriving insights and facilitating data-driven decisions. Within topic analysis, the primary machine learning techniques employed areas of focus include topic modeling and topic classification within this field. However, topic modeling encounters various challenges, specific to document properties. NLP is an integrative subject that merges CS, AI, and linguistics to construct systems capable of comprehending and processing human language. The prevalent machine using labeled data to categorize unlabeled data. This process relies on the knowledge gained during training to classify new data. In general, text classification methods handle predefined and finite categories such as predicting labels like credible or not credible for credibility assessment, or determining movie ratings (bad, okay, good) based on reviews. The difficulty in text classification arises from the predetermined set of topics or labels. When the topics are not known in advance, the concept of topic modeling becomes crucial. This statistical modeling approach is designed to identify abstract topics within a set of documents that lack predefined labels. By analyzing labelled data, this method extracts underlying topics.

In this era, technology has significantly simplified people’s lives, and one of the recent advancements in artificial intelligence is deep learning. Deep learning has emerged as a field that enables the creation of intelligent software and machines capable of assisting individuals in their daily tasks. One notable application of deep learning is Human Activity Recognition (HAR). Deep learning, a subset of machine learning, is used effectively to identify human activities. In this project, we used a model based on Convolutional Long Short-Term Memory (ConvLSTM) and Long-term Recurrent Convolutional Network (LRCN) to detect human activities. This model is trained on the UCF50 dataset, which allows rigorous testing and validation. A dataset is created from the main dataset (UCF 50) with 10 action categories, and further, the dataset is split into two parts: testing and validation. Using the subsequent dataset, the ConvLSTM model accuracy is 81.4%, and the LRCN model accuracy is 85.3%.

Among the most rapidly developing areas in computer vision is object detection. Mask detection is the main objective of the effort. With the use of deep learning and computer vision techniques, this project offers a reliable method for mask identification that is implemented using RESNET architecture. Identifying faces and differentiating between people wearing masks and those without is the main goal. The model is refined via transfer learning on a customized dataset that includes annotated photos of faces that have been masked, masked incorrectly and unmasked faces.

BACKGROUND: Heart disease represents the leading cause of death globally.Timely diagnosis and treatment can prevent cardiovascular issues.An Electrocardiograms (ECG) serves as a diagnostic tool for identifying heart difficulties.Cardiovascular Disease (CVD) often gets identified through ECGs.Deep learning (DL) garners attention in healthcare due to its potential in swiftly diagnosing ECG anomalies, crucial for patient monitoring.Conversely, automatic CVD detection from ECGs poses a challenging task, wherein rule-based diagnostic models usually achieve top-notch performance.These models encounter complications in supervision vast volumes of diverse data, demanding widespread analysis and medical capability to ensure precise CVD diagnosis.OBJECTIVE: This study aims to enhance cardiovascular disease diagnosis by combining symptom-based detection and ECG analysis.METHODS: To enhance these experiments, we built a novel automated prediction method based on a Feed Forward Neural Network (FFNN) model.The fundamental objective of our method is to develop the accuracy of ECG diagnosis.Our strategy employs chaos theory and destruction analysis to combine optimum deep learning features with a well-organized set of ECG properties.In addition, we use the constant-Q non-stationary Gabor transform (CQNGT) to convert one-dimensional ECG data into a two-dimensional picture.A pre-trained FFNN processes this image.To identify significant features from the FFNN output that correspond with the ECG data, we employ pairwise feature proximity.RESULTS: According to experimental findings, the suggested system, FFNN-CQNGT, surpasses other state-of-the-art systems in terms of precision of 94.89%, computational efficiency of 2.114 ms, accuracy of 95.55%, specificity of 93.77%, and sensitivity of 93.99% and MSE 40.32%.CONCLUSION: Contributing an automated ECG-based DL system based on FFNN-CQNGT for early-stage cardiovascular disease identification and classification holds great potential for both patient care and public health.

In this study, we investigate how AI and ML might revolutionize the agricultural industry, particularly with regard to increasing crop output while decreasing input costs. Applying AI and ML technology has promise in a society struggling with population increase, climate change, and resource constraints. This study highlights the practical advantages of AI and ML in agriculture via a well-crafted research process, including data gathering, model creation, and assessment. The results show that AI and ML models are useful for forecasting agricultural yields, identifying illnesses, allocating resources efficiently, and assisting farmers with decision-making based on empirical evidence. Results like this highlight the importance of these technologies in advancing goals of efficiency, sustainability, and food safety. Additionally, the study acknowledges the significance of addressing ethical problems in AI deployment, guaranteeing equal access to these advancements. We should expect to see more research into cutting-edge methods, Internet of Things (IoT) integration, and accessible tools for subsistence farmers as we go further in the use of AI and ML in the agricultural sector. The full promise of AI and ML in designing a resilient, productive, and sustainable agricultural future requires collaborative efforts across stakeholders. In the struggle to feed the globe while protecting its resources, this study shines a bright light of optimism.

Remote underwater picture and video capture is used by marine biologists to monitor different fish species in their natural environments. This aids in their comprehension and forecasting of the responses of fish to fishing pressure, habitat degradation, and climate change. Having this knowledge is crucial for creating environmentally friendly, sustainable fisheries for human use. Humans, on the other hand, find it difficult and time-consuming to extract useful information from massive amounts of collected videos. Deep learning (DL) appears to have an issue with this. With the help of DL, marine biologists can rapidly and effectively parse massive amounts of film, uncovering specialized information that is not accessible via manual monitoring techniques. We present a two-step deep learning technique in this study that can recognize and classify temperate fishes without the use of pre-filtering. Every fish in a picture must first be identified, regardless of species or gender. For this, we employ the You Only Look Once (YOLO) object detection technique. The classification of each fish in the image is done in the second stage using a squeeze-and-excitation (SE)-designed convolutional neural network (CNN). Despite the short training sample size of temperate fishes, we use transfer learning to improve classification accuracy. For this, the fish classifier was trained using a public dataset, and the object detection model was trained using ImageNet. Both models were then updated with pertinent temperate fishes. Weights are always added both before and after a workout. The CNN-SE model performed admirably, with a 96.22% accuracy. Extensive comparative research revealed that the CNN-SE model outperformed more recent approaches.

Crop yield is an extremely difficult trait identified using many factors like genotype, environment and their interaction. Accurate Crop Yield Prediction (CYP) necessitates the basic understanding of the functional relativity among yields and the collaborative factor. Disclosing such connection requires both wide-ranging datasets and an efficient model. The CYP is important to accomplish irrigation scheduling and assessing labor necessities for reaping and storing. Predicting yield using various kinds of irrigation is effective for optimizing resources, but CYP is a difficult process owing to the existence of distinct factors. Recently, Deep Learning (DL) approaches offer solutions to complicated data like weather parameters, maturity groups, genotype, etc. In this aspect, this paper presents an Automated Crop Yield Prediction utilizing Chaotic Political Optimizer with Deep Learning (ACYP-CPODL) model. The proposed ACYP-CPODL technique involves different processes namely pre-processing, prediction and parameter optimization. In addition, the hybrid Convolutional Neural Network (CNN) Long-Short Term Memory (LSTM) technique is designed for the prediction process. Moreover, the hyperparameter tuning of the CNN-LSTM approach is performed by the CPO algorithm. The proposed ACYP-CPODL technique has produced an effective result with an MSE of 0.031 and R2 Score of 0.936, whereas the BLSTM model has produced a near-optimal results. As a result, the proposed ACYP-CPODL method has proven to be an effective tool for predicting the crop yields. For validating the improved predictive performance of the ACYP-CPODL technique, a wide range of simulations take place on benchmark datasets and the comparative results highlighted the betterment of the ACYP-CPODL technique over the recent methods.

Several studies aimed at improving healthcare management have shown that the importance of healthcare has grown in recent years. In the healthcare industry, effective decision-making requires multicriteria group decision-making. Simultaneously, big data analytics could be used to help with disease detection and healthcare delivery. Only a few previous studies on large-scale group decision-making (LSDGM) in the big data-driven healthcare Industry 4.0 have focused on this topic. The goal of this work is to improve healthcare management decision-making by developing a new MapReduce-based LSDGM model (MR-LSDGM) for the healthcare Industry 4.0 context. Clustering decision-makers (DM), modelling DM preferences, and classification are the three stages of the MR-LSDGM technique. Furthermore, the DMs are subdivided using a novel biogeography-based optimization (BBO) technique combined with fuzzy C-means (FCM). The subgroup preferences are then modelled using the two-tuple fuzzy linguistic representation (2TFLR) technique. The final classification method also includes a feature extractor based on long short-term memory (LSTM) and a classifier based on an ideal extreme learning machine (ELM). MapReduce is a data management platform used to handle massive amounts of data. A thorough set of experimental analyses is carried out, and the results are analysed using a variety of metrics.

Advancements in recent networking and information technology have always been a natural phenomenon. The exponential amount of data generated by the people in their day-to-day lives results in the rise of Big Data Analytics (BDA). Cognitive computing is an Artificial Intelligence (AI) based system that can reduce the issues faced during BDA. On the other hand, Sentiment Analysis (SA) is employed to understand such linguistic based tweets, feature extraction, compute subjectivity and sentimental texts placed in these tweets. The application of SA on big data finds it useful for businesses to take commercial benefits insight from text-oriented content. In this view, this paper presents new cognitive computing with the big data analysis tool for SA. The proposed model involves various process such as pre-processing, feature extraction, feature selection and classification. For handling big data, Hadoop Map Reduce tool is used. The proposed model initially undergoes pre-processing to remove the unwanted words. Then, Term Frequency-Inverse Document Frequency (TF-IDF) is utilized as a feature extraction technique to extract the set of feature vectors. Besides, a Binary Brain Storm Optimization (BBSO) algorithm is being used for the Feature Selection (FS) process and thereby achieving improved classification performance. Moreover, Fuzzy Cognitive Maps (FCMs) are used as a classifier to classify the incidence of positive or negative sentiments. A comprehensive experimental results analysis ensures the better performance of the presented BBSO-FCM model on the benchmark dataset. The obtained experimental values highlights the improved classification performance of the proposed BBSO-FCM model in terms of different measures.

Over the last few years, electric cars (EVs) have grown in popularity. The battery management system (BMS) is critical to the long-term viability and smooth operation of an electric vehicle. The management of electric vehicles' batteries daily may help them operate better. All battery-related data is monitored and transmitted to the cloud in real time for monitoring via the Internet of Things, which is completely automated. Open loop approaches for predicting SOC, SOH, and SOP parameters suffer from a decrease in reliability when current sensor uncertainties and the rarity of relaxation status are addressed. Also, keep in mind that IoT network nodes are generally delay-tolerant, and message delivery latency still has a substantial impact on monitoring an electric vehicle's battery system. (EV). The research team created an Internet of Things BMS based on LR parameter estimation and an ORMeshNet gateway topology to address this issue. Before any other systems can be monitored or diagnosed with the BMS or any other system, techniques are first created based on an LR to estimate SOC and SOH accurately and efficiently. This method achieves a higher rate of convergence as well as a higher level of fault tolerance than other estimation methods. The updated parameters and estimated states are used for the SOP estimator to provide more accurate peak power estimates while fulfilling operational constraints of the battery current, voltage, and SOC. Thereafter, the estimated results are transferred via IoT platform that comprises of OTH-AJS node selection followed with LND-BES optimal routing based MeshNet gateway protocol to transfer the data for monitoring. The proposed approach yields a throughput of 88.97%, a PDR of 87.98%, and a Goodput of 83.98%. Experimental results show enormous improvement in estimating the parameters with better throughput, PDR, and goodput value as compared to existing methods.

Magnetic resonance image (MRI) is one of major component in medical brain image, imaging technique and segmentation of brain medical image is a crucial & complex task in evaluation of MRI images. Conventionally, different types of fuzzy, soft set related approaches like intuitionistic, fuzzy c-means, fuzzy c-means were developed to segmentation of brain related image, but these approaches face accuracy loss in brain image segmentation. So we consider new segmentation approach i.e. Intuitionistic neutro soft based rough sets and Classical Regression model (INSRCRM) which is extension to Advanced machine learning approach i.e. Enhanced & Explored Intuitionistic FCM clustering (EEISFCM) for smoothness and to increase image accuracy and intensity. Proposed approach is applied to increase accuracy and intensity with respect to spatial data processing for medical brain image segmentation and evaluate histon and histogram based image smoothness. Proposed approach evaluated with lower and upper approximations for intensity based brain image segmentation. This approach mainly identifies real valleys to smooth measure to present brain image segmentation to reduce noise reduction based on threshold of image pixels with different image notations. Experimental results of proposed approach gives to find peaks and valleys to demonstrate better image segmentation results with respect to traditional approaches.

Intuitionistic fuzzy sets (IFSs) and rough sets are extensively used mathematical tools to handle uncertainty and vagueness present in images and recently are combined together to segment MR medical images in the presence of intensity non uniformity (INU) and noise. In this paper, a novel clustering algorithm, namely modified rough intuitionistic fuzzy c-means (MRIFCM) is proposed for the segmentation of the brain magnetic resonance images to extract the white matter, grey matter and the cerebrospinal fluid from MR brain image with bias field correction. A new intuitionistic fuzzy complement function is proposed for intuitionistic fuzzy image representation to take into account intensity in homogeneity and noise in brain MR images.Further, Hausdorff distance is used as distance metric to calculate the distance between cluster center and pixel. The proposed algorithm is evaluated through simulation and compared it with existing k-means (KM),Rough k-means(RKM), fuzzy C-means (FCM), Rough fuzzy c-means(RFCM), Generalized rough fuzzy c-means (GRFCM), soft fuzzy rough c-means (SFRCM),rough intuitionistic fuzzy c-means(RIFCM) and Generalized rough Intuitionistic fuzzy c-means(GRIFCM) algorithms. Experimental results prove the superiority of the proposed algorithm over the considered algorithms in all analyzed scenarios.