Search

In response to the growing number of diabetes cases worldwide, Our study addresses the escalating issue of diabetic eye disease (DED), a significant contributor to vision loss globally, through a pioneering approach. We propose a novel integration of a Genetic Grey Wolf Optimization (G-GWO) algorithm with a Fully Convolutional Encoder-Decoder Network (FCEDN), further enhanced by a Kernel Extreme Learning Machine (KELM) for refined image segmentation and disease classification. This innovative combination leverages the genetic algorithm and grey wolf optimization to boost the FCEDN's efficiency, enabling precise detection of DED stages and differentiation among disease types. Tested across diverse datasets, including IDRiD, DR-HAGIS, and ODIR, our model showcased superior performance, achieving classification accuracies between 98.5% to 98.8%, surpassing existing methods. This advancement sets a new standard in DED detection and offers significant potential for automating fundus image analysis, reducing reliance on manual examination, and improving patient care efficiency. Our findings are crucial to enhancing diagnostic accuracy and patient outcomes in DED management., Competing Interests: The authors have declared that no competing interests exist., (Copyright: © 2024 Khan et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.)

Cardiotocography measures the fetal heart rate in the fetus during pregnancy to ensure physical health because cardiotocography gives data about fetal heart rate and uterine shrinkages which is very beneficial to detect whether the fetus is normal or suspect or pathologic. Various cardiotocography measures infer wrongly and give wrong predictions because of human error. The traditional way of reading the cardiotocography measures is the time taken and belongs to numerous human errors as well. Fetal condition is very important to measure at numerous stages and give proper medications to the fetus for its well-being. In the current period Machine learning (ML) is a well-known classification strategy used in the biomedical field on various issues because ML is very fast and gives appropriate results that are better than traditional results. ML techniques play a pivotal role in detecting fetal disease in its early stages. This research article uses Federated machine learning (FML) and ML techniques to classify the condition of the fetus. This study proposed a model for the detection of bio-signal cardiotocography that uses FML and ML techniques to train and test the data. So, the proposed model of FML used numerous data preprocessing techniques to overcome data deficiency and achieves 99.06% and 0.94% of prediction accuracy and misprediction rate, respectively, and parallel the proposed model applying K-nearest neighbor (KNN) and achieves 82.93% and 17.07% of prediction accuracy and misprediction accuracy, respectively. So, by comparing both models FML outperformed the KNN technique and achieved the best and most appropriate prediction results as compared with previous studies the proposed study achieves the best and most accurate results. [ABSTRACT FROM AUTHOR]

Network operations involve several decision-making tasks. Some of these tasks are related to operators, such as extending the footprint or upgrading the network capacity. Other decision tasks are related to network functions, such as traffic classifications, scheduling, capacity, coverage trade-offs, and policy enforcement. These decisions are often decentralized, and each network node makes its own decisions based on the preconfigured rules or policies. To ensure effectiveness, it is essential that planning and functional decisions are in harmony. However, human intervention-based decisions are subject to high costs, delays, and mistakes. On the other hand, machine learning has been used in different fields of life to automate decision processes intelligently. Similarly, future intelligent networks are also expected to see an intense use of machine learning and artificial intelligence techniques for functional and operational automation. This article investigates the current state-of-the-art methods for packet scheduling and related decision processes. Furthermore, it proposes a machine learningbased approach for packet scheduling for agile and cost-effective networks to address various issues and challenges. The analysis of the experimental results shows that the proposed deep learning-based approach can successfully address the challenges without compromising the network performance. For example, it has been seen that with mean absolute error from 6.38 to 8.41 using the proposed deep learning model, the packet scheduling can maintain 99.95% throughput, 99.97% delay, and 99.94% jitter, which are much better as compared to the statically configured traffic profiles. [ABSTRACT FROM AUTHOR]

The Internet of Things (IoT) has emerged as a fundamental framework for interconnected device communication, representing a relatively new paradigm and the evolution of the Internet into its next phase. Its significance is pronounced in diverse fields, especially healthcare, where it finds applications in scenarios such as medical service tracking. By analyzing patterns in observed parameters, the anticipation of disease types becomes feasible. Stress monitoring with wearable sensors and the Internet of Things (IoT) is a potential application that can enhance wellness and preventative health management. Healthcare professionals have harnessed robust systems incorporating battery-based wearable technology and wireless communication channels to enable cost-effective healthcare monitoring for various medical conditions. Network-connected sensors, whether within living spaces or worn on the body, accumulate data crucial for evaluating patients' health. The integration of machine learning and cutting-edge technology has sparked research interest in addressing stress levels. Psychological stress significantly impacts a person's physiological parameters. Stress can have negative impacts over time, prompting sometimes costly therapies. Acute stress levels can even constitute a life-threatening risk, especially in people who have previously been diagnosed with borderline personality disorder or schizophrenia. To offer a proactive solution within the realm of smart healthcare, this article introduces a novel machine learning-based system termed "Stress-Track". The device is intended to track a person's stress levels by examining their body temperature, sweat, and motion rate during physical activity. The proposed model achieves an impressive accuracy rate of 99.5%, showcasing its potential impact on stress management and healthcare enhancement. [ABSTRACT FROM AUTHOR]

Objective: The endoscopic and laryngoscopic examination is paramount for laryngeal, oropharyngeal, nasopharyngeal, nasal, and oral cavity benign lesions and cancer evaluation. Nevertheless, upper aerodigestive tract (UADT) endoscopy is intrinsically operator‐dependent and lacks objective quality standards. At present, there has been an increased interest in artificial intelligence (AI) applications in this area to support physicians during the examination, thus enhancing diagnostic performances. The relative novelty of this research field poses a challenge both for the reviewers and readers as clinicians often lack a specific technical background. Data Sources: Four bibliographic databases were searched: PubMed, EMBASE, Cochrane, and Google Scholar. Review Methods: A structured review of the current literature (up to September 2022) was performed. Search terms related to topics of AI, machine learning (ML), and deep learning (DL) in UADT endoscopy and laryngoscopy were identified and queried by 3 independent reviewers. Citations of selected studies were also evaluated to ensure comprehensiveness. Conclusions: Forty‐one studies were included in the review. AI and computer vision techniques were used to achieve 3 fundamental tasks in this field: classification, detection, and segmentation. All papers were summarized and reviewed. Implications for Practice: This article comprehensively reviews the latest developments in the application of ML and DL in UADT endoscopy and laryngoscopy, as well as their future clinical implications. The technical basis of AI is also explained, providing guidance for nonexpert readers to allow critical appraisal of the evaluation metrics and the most relevant quality requirements. [ABSTRACT FROM AUTHOR]

Citrus is manually counted to estimate the yield. By using some innovative agricultural techniques yield and production can be increased. Numerous agricultural innovations have been introduced in recent years. Higher agricultural production, prediction, and reliable crop status information are more important than ever due to the expected growth of the human population. Agriculture has always been the foundation of human society. Current study was aimed to develop a reliable and meaningful information-gathering agricultural field based on image processing during 2020. Citrus yield can be increased in the initial stages by counting it with RGB and HSV-based images taken from an Android phone from various angles using machine learning techniques. Fertilizers such as potash, phosphorus, and nitrogen can then be utilized to boost yield. According to the findings, farmers can control and monitor citrus health production more efficiently and effectively by integrating machine learning with agriculture. The citrus calculation using the given technique compared with manually counted citrus, having difference of up to 5 to 10 citruses for a single plant per plot in a field. The proposed method produced excellent results under varying lighting conditions, leaf occlusion, and fruit overlap on photos taken at various distances from the orange trees. [ABSTRACT FROM AUTHOR]

Introduction: Narrow Band Imaging (NBI) is an endoscopic visualization technique useful for upper aero-digestive tract (UADT) cancer detection and margins evaluation. However, NBI analysis is strongly operator-dependent and requires high expertise, thus limiting its wider implementation. Recently, artificial intelligence (AI) has demonstrated potential for applications in UADT videoendoscopy. Among AI methods, deep learning algorithms, and especially convolutional neural networks (CNNs), are particularly suitable for delineating cancers on videoendoscopy. This study is aimed to develop a CNN for automatic semantic segmentation of UADT cancer on endoscopic images. Materials and Methods: A dataset of white light and NBI videoframes of laryngeal squamous cell carcinoma (LSCC) was collected and manually annotated. A novel DL segmentation model (SegMENT) was designed. SegMENT relies on DeepLabV3+ CNN architecture, modified using Xception as a backbone and incorporating ensemble features from other CNNs. The performance of SegMENT was compared to state-of-the-art CNNs (UNet , ResUNet , and DeepLabv3). SegMENT was then validated on two external datasets of NBI images of oropharyngeal (OPSCC) and oral cavity SCC (OSCC) obtained from a previously published study. The impact of in-domain transfer learning through an ensemble technique was evaluated on the external datasets. Results: 219 LSCC patients were retrospectively included in the study. A total of 683 videoframes composed the LSCC dataset, while the external validation cohorts of OPSCC and OCSCC contained 116 and 102 images. On the LSCC dataset, SegMENT outperformed the other DL models, obtaining the following median values: 0.68 intersection over union (IoU), 0.81 dice similarity coefficient (DSC), 0.95 recall, 0.78 precision, 0.97 accuracy. For the OCSCC and OPSCC datasets, results were superior compared to previously published data: the median performance metrics were, respectively, improved as follows: DSC=10.3% and 11.9%, recall=15.0% and 5.1%, precision=17.0% and 14.7%, accuracy=4.1% and 10.3%. Conclusion: SegMENT achieved promising performances, showing that automatic tumor segmentation in endoscopic images is feasible even within the highly heterogeneous and complex UADT environment. SegMENT outperformed the previously published results on the external validation cohorts. The model demonstrated potential for improved detection of early tumors, more precise biopsies, and better selection of resection margins. [ABSTRACT FROM AUTHOR]

Sentiment analysis is a method for assessing an individual's thought, opinion, feeling, mentality, and conviction about a specific subject on indicated theme, idea, or product. The point could be a business association, a news article, a research paper, or an online item, etc. Opinions are generally divided into three groups of positive, negative, and unbiased. The way toward investigating different opinions and gathering them in every one of these categories is known as Sentiment Analysis. The enormously growing sentiment data on the web especially social media can be a big source of information. The processing of this unstructured data is of deep interest for researchers in the field of data science. The proposed article is an effort to find a model which can improve the classification accuracy of sentiment data. In this paper, a complete model is introduced to improve the accuracy of binary-class sentiment data. The model is reliable as it has been validated on three distinctive datasets with various sample sizes. In this regard, Amazon reviews, Yelp reviews, and IMDB reviews are taken into account. This model is based on completely referenced datasets and considerable improvement in accuracy is seen as compared to individual classifiers. Various heterogeneous classifiers are trained on the above-mentioned datasets at the base level. These base-level classifiers ought to be chosen from various groups of classifiers. At that point, the output of these base-level classifiers is consolidated to create a new dataset that acts as input for meta-level classifiers. At the meta-level, the base-level classifier with the best outcomes is selected with the new dataset framed. The model accordingly framed can be used to predict any sample query. [ABSTRACT FROM AUTHOR]

Stock market trends can be affected by external factors such as public sentiment and political events. The goal of this research is to find whether or not public sentiment and political situation on a given day can affect stock market trends of individual companies or the overall market. For this purpose, the sentiment and situation features are used in a machine learning model to find the effect of public sentiment and political situation on the prediction accuracy of algorithms for 7 days in future. Besides, interdependencies among companies and stock markets are also studied. For the sake of experimentation, stock market historical data are downloaded from Yahoo! Finance and public sentiments are obtained from Twitter. Important political events data of Pakistan are crawled from Wikipedia. The raw text data are then pre-processed, and the sentiment and situation features are generated to create the final data sets. Ten machine learning algorithms are applied to the final data sets to predict the stock market future trend. The experimental results show that the sentiment feature improves the prediction accuracy of machine learning algorithms by 0–3%, and political situation feature improves the prediction accuracy of algorithms by about 20%. Furthermore, the sentiment attribute is most effective on day 7, while the political situation attribute is most effective on day 5. SMO algorithm is found to show the best performance, while ASC and Bagging show poor performance. The interdependency results indicate that stock markets in the same industry show a medium positive correlation with each other. [ABSTRACT FROM AUTHOR]

The continuous progress in wearable sensing technologies has motivated the researchers to develop novel models for human activity and behavior monitoring. As wearable sensors possess more liberty in their placement at multiple positions on the user's body to track human motion patterns, hence, they have been extensively utilized in activity recognition systems. However, wearable inertial sensors are prone to their position and orientation sensitivity, thus leading to poor recognition performance in real-time scenarios. Therefore, in this study, we address the problem of position-independent human activity recognition using the wearable sensor. In this aspect, we propose a set of linear and non-linear transformations for 3D-sensor data to minimize the position and orientation sensitivity of the inertial sensor. We also present a feature extraction framework to efficiently recognize human activities independent of any sensor position. Finally, we validate our proposed scheme using the PAMAP dataset, which achieves the best average performance of 94.7% and 91.7% for position-dependent and position-independent activity recognition. [ABSTRACT FROM AUTHOR]

We live in a world where ubiquitous systems surround us in the form of automated homes, smart appliances and wearable devices. These ubiquitous systems not only enhance productivity but can also provide assistance given a variety of different scenarios. However, these systems are vulnerable to the risk of unauthorized access, hence the ability to authenticate the end-user seamlessly and securely is important. This paper presents an approach for user identification given the physical activity patterns captured using on-body wearable sensors, such as accelerometer, gyroscope, and magnetometer. Three machine learning classifiers have been used to discover the activity patterns of users given the data captured from wearable sensors. The recognition results prove that the proposed scheme can effectively recognize a user's identity based on his/her daily living physical activity patterns. [ABSTRACT FROM AUTHOR]

Precise and reliable hydrological runoff prediction plays a significant role in the optimal management of hydropower resources. Nevertheless, the hydrological runoff practically possesses a nonlinear dynamics, and constructing appropriate runoff prediction models to deal with the nonlinearity is a challenging task. To overcome this difficulty, this paper proposes a three-stage novel hybrid model, namely, CVS (CEEMDAN-VMD-SVM), by coupling the support vector machine (SVM) with a two-stage signal decomposition methodology, combining complete ensemble empirical decomposition with additive noise (CEEMDAN) and variational mode decomposition (VMD), to obtain inclusive information of the runoff time series. Hydrological runoff data of the Swat River, Pakistan, from 1961 to 2015 were taken for prediction. CEEMDAN decomposes the runoff time series into subcomponents, and VMD performs further decomposition of the high-frequency component obtained after CEEMDAN decomposition to improve the prediction activity. Afterward, the SVM algorithm was applied to the decomposed subcomponents for the prediction purpose. Finally, four statistical indices are utilized to measure the performance of the CVS model compared with other hybrid models including CEEMDAN-VMD-MLP (multilayer perceptron), CEEMDAN-SVM, VMD-SVM, CEEMDAN-MLP, VMD-MLP, SVM, and MLP. The CVS model performs better during the training period by reducing RMSE by 71.28% and 40.06% compared with MLP and CEEDMAD-VMD-SVM models, respectively. However, during the testing period, the error reductions include RMSE by 68.37% and 35.33% compared with MLP and CEEDMAD-VMD-SVM models, respectively. The results highlight that the CVS model outperforms other models in terms of accuracy and error reduction. The research also highlights the superiority of other hybrid models over standalone in predicting the hydrological runoff. Therefore, the proposed hybrid model is applicable for the nonlinear features of runoff time series with feasibility for future planning and management of water resources. [ABSTRACT FROM AUTHOR]

Falling is a severe hazard among older adults. Fall treatment is considered to be one of the most costly treatments, which usually extends to a long time. One bad fall can cause severe injuries that may lead to permanent disability or even death. Therefore, an efficient and cost-effective fall monitoring system is exceptionally indispensable. With the advancement in technology, wearable sensors and systems provide a lucrative way to continuously monitor the elderly people for detecting any fall incident that may occur. Most of these wearable fall monitoring systems focus only on detecting a fall incident. However, to avoid the risk of any future fall, it is essential to be aware of the cause of a fall incident also. Therefore, to address this challenge, a wearable sensor-based continuous fall monitoring system is proposed in this paper, which is capable of detecting a fall and identifying the falling pattern and the activity associated with the fall incident. The performance of the proposed scheme is investigated with a series of experiments using three machine learning algorithms, namely, $k$ -nearest neighbors (KNNs), support vector machine, and random forest (RF). The proposed methodology achieved the highest accuracy for fall detection, i.e., 99.80%, using KNNs classifier, whereas the highest accuracy achieved in recognizing different falling activities is 96.82% using RF classifier. [ABSTRACT FROM AUTHOR]

Machine learning (ML), a branch of artificial intelligence (AI), is an area of computational science that is concerned with the analysis and interpretation of patterns and structures in data to enable learning and decision-making without the participation of a human. Hidden Markov models (HMMs), which have been acknowledged for decades but have recently made a significant revival in machine learning, are one of the most impressively powerful probabilistic models. HMMs are frequently employed in machine learning to model heterogeneous time series data. In this paper, we integrate independent component analysis (ICA) and ICA with a bounded multivariate generalized Gaussian mixture model (ICA-BMGGMM) into the HMM approach. One limitation of ICA is that it assumes the sources to be independent from each other. This assumption can be relaxed by combining independent vectors analysis (IVA) and IVA with the BMGGMM (IVA-BMGGMM) into the HMM approach to improve their modeling capability. We validate our proposed models using a variety of applications, such as human action recognition, speech recognition, and energy disaggregation. The results presented in the paper demonstrate the effectiveness of the proposed approaches for modeling different types of data. These data include KTH and Weizmann datasets for human action recognition, TIMIT and SDR for speech recognition, REDD dataset for energy disaggregation and EEG dataset for elliptic seizure classification. For all conducted experiments, our proposed models outperform other comparing models for all performance metrics such as accuracy, sensitivity, and precision. The best detection results were found using the IVABMGGMM-HMM for the reported experiments. [ABSTRACT FROM AUTHOR]

Classification trees are top-down induction of labeled sampling units into recursive order to get end nodes. Each end node representing those labeled units, which are in majority, otherwise considered as misclassified. In the top-down induction process, an evaluation criterion plays an important role to send maximum of the units having same label to the same node. To achieve this goal a "goodness of split" measure is calculated by using evaluation criteria e.g. Gini function, Twoing rule, Entropy function etc. for each distinct value of each variable and finally chooses one, which enhances the purity. Almost all the evaluation criteria provide the same results in terms of misclassified units, deviance and number of end nodes. Here we obtain simulation results on the proposed evaluation criteria (Azam, et al., 2007) to split a node, which fulfil all the required properties of any evaluation criteria (Breiman et al., 1984) and also propose a top-down sequential pruning scheme. We conducted a simulation study to test the performance of the proposed evaluation criterion using the top-down sequential pruning scheme over many real life datasets available under UCI Machine Learning Repository and observed that the proposed strategy provides improved results. [ABSTRACT FROM AUTHOR]

The optimal management of hydropower resources is highly dependent on accurate and reliable hydrological runoff forecasting. The development of a suitable runoff-forecasting model is a challenging task due to the complex and nonlinear nature of runoff. To meet the challenge, this study proposed a three-stage novel hybrid model namely IVG (ICEEMDAN-VMD-GRU), by coupling gated recurrent unit (GRU) with a two-stage signal decomposition methodology, combining improved complete ensemble empirical decomposition with additive noise (ICEEMDAN) and variational mode decomposition (VMD), to forecast the monthly runoff of SWAT river, Pakistan. ICEEMDAN decomposed the runoff time series into subcomponents, and VMD performed further decomposition of the high-frequency component obtained by ICEEMDAN decomposition. Afterward, the GRU network was employed to the decomposed subcomponents for forecasting purposes. The performance of the IVG model was compared with other hybrid models including, ICEEMDAN-VMD-SVM (support vector machine), ICEEMDANGRU, VMD-GRU, ICEEMDAN-SVM, VMD-SVM; and standalone models including GRU and SVM by utilizing statistical indices. Experimental results proved that the IVG model outperformed other models in terms of accuracy and error reduction, which indicates the feasibility of the IVG model to analyze the nonlinear features of runoff time series and for runoff forecasting with applicability for future planning and management of water resources. [ABSTRACT FROM AUTHOR]

Personalized federated learning (PFL) is an exciting approach that allows machine learning (ML) models to be trained on diverse and decentralized sources of data, while maintaining client privacy and autonomy. However, PFL faces several challenges that can deteriorate the performance and effectiveness of the learning process. These challenges include data heterogeneity, communication overhead, model privacy, model drift, client heterogeneity, label noise and imbalance, federated optimization challenges, and client participation and engagement. To address these challenges, researchers are exploring innovative techniques and algorithms that can enable efficient and effective PFL. These techniques include several optimization algorithms. This research survey provides an overview of the challenges and motivations related to the model optimization strategies for PFL, as well as the state-of-the-art (SOTA) methods and algorithms which seek to provide solutions of these challenges. Overall, this survey can be a valuable resource for researchers who are interested in the emerging field of PFL as well as its potential for personalized machine learning in a federated environment. • A Comprehensive Review of Contemporary Approaches in Personalized Federated Learning (PFL). • Categorization and classification of various methods and solutions to provide comprehensive taxonomies. • A diverse array of literature that addresses emerging challenges and cutting-edge solutions in the field. • This paper outlines and proposes new research directions that can be pursued by the interested community. [ABSTRACT FROM AUTHOR]