Balanced Dataset Research Articles

The Role of Machine Learning in Enhancing Particulate Matter Estimation: A Systematic Literature Review

As urbanization and industrial activities accelerate globally, air quality has become a pressing concern, particularly due to the harmful effects of particulate matter (PM), notably PM2.5 and PM10. This review paper presents a comprehensive systematic assessment of machine learning (ML) techniques for estimating PM concentrations, drawing on studies published from 2018 to 2024. Traditional statistical methods often fail to account for the complex dynamics of air pollution, leading to inaccurate predictions, especially during peak pollution events. In contrast, ML approaches have emerged as powerful tools that leverage large datasets to capture nonlinear, intricate relationships among various environmental, meteorological, and anthropogenic factors. This review synthesizes findings from 32 studies, demonstrating that ML techniques, particularly ensemble learning models, significantly enhance estimation accuracy. However, challenges remain, including data quality, the need for diverse and balanced datasets, issues related to feature selection, and spatial discontinuity. This paper identifies critical research gaps and proposes future directions to improve model robustness and applicability. By advancing the understanding of ML applications in air quality monitoring, this review seeks to contribute to developing effective strategies for mitigating air pollution and protecting public health.

Theoretical framework for the difference of two negative binomial distributions and its application in comparative analysis of sequencing data.

High-throughput sequencing (HTS) technologies have been instrumental in investigating biological questions at the bulk and single-cell levels. Comparative analysis of two HTS data sets often relies on testing the statistical significance for the difference of two negative binomial distributions (DOTNB). Although negative binomial distributions are well studied, the theoretical results for DOTNB remain largely unexplored. Here, we derive basic analytical results for DOTNB and examine its asymptotic properties. As a state-of-the-art application of DOTNB, we introduce DEGage, a computational method for detecting differentially expressed genes (DEGs) in scRNA-seq data. DEGage calculates the mean of the sample-wise differences of gene expression levels as the test statistic and determines significant differential expression by computing the P-value with DOTNB. Extensive validation using simulated and real scRNA-seq data sets demonstrates that DEGage outperforms five popular DEG analysis tools: DEGseq2, DEsingle, edgeR, Monocle3, and scDD. DEGage is robust against high dropout levels and exhibits superior sensitivity when applied to balanced and imbalanced data sets, even with small sample sizes. We utilize DEGage to analyze prostate cancer scRNA-seq data sets and identify marker genes for 17 cell types. Furthermore, we apply DEGage to scRNA-seq data sets of mouse neurons with and without fear memory and reveal eight potential memory-related genes overlooked in previous analyses. The theoretical results and supporting software for DOTNB can be widely applied to comparative analyses of dispersed count data in HTS and broad research questions.

Enhancing the accuracy of blockage detection in centrifugal pump using the majority voting classifier on an unbalanced dataset

Blockage in a centrifugal pump can adversely impact its performance. In this study, an experimental facility is developed to simulate three types of blockages (suction, discharge, and simultaneous suction and discharge). To classify these faults and detect their severity, a methodology involving the application of majority voting classifier (MVC) to the pump’s discharge pressure signals is presented. An unbalanced dataset is constructed, where the number of samples for a specific blockage condition decreases with increasing fault severity. Statistical features, entropy features, and entropies meta features are extracted from the signal and ranked using XGBoost and minimum redundancy maximum relevance (MRMR). Subsequently, the optimal features are selected based on the best performing model (Linear Discriminant Analysis) among ten different models. Best four models are selected and ensembled to form MVC. Results show that MVC achieves an accuracy of 89.90% and 88.26% for features selected by XGBoost and MRMR, respectively. Finally, the unbalanced dataset is balanced using synthetic minority oversampling and it is shown that MVC achieves an accuracy of 100% on this balanced dataset for the features selected using both approaches.

ACGAN for Addressing the Security Challenges in IoT-Based Healthcare System

The continuous evolution of the IoT paradigm has been extensively applied across various application domains, including air traffic control, education, healthcare, agriculture, transportation, smart home appliances, and others. Our primary focus revolves around exploring the applications of IoT, particularly within healthcare, where it assumes a pivotal role in facilitating secure and real-time remote patient-monitoring systems. This innovation aims to enhance the quality of service and ultimately improve people’s lives. A key component in this ecosystem is the Healthcare Monitoring System (HMS), a technology-based framework designed to continuously monitor and manage patient and healthcare provider data in real time. This system integrates various components, such as software, medical devices, and processes, aimed at improvi1g patient care and supporting healthcare providers in making well-informed decisions. This fosters proactive healthcare management and enables timely interventions when needed. However, data transmission in these systems poses significant security threats during the transfer process, as malicious actors may attempt to breach security protocols.This jeopardizes the integrity of the Internet of Medical Things (IoMT) and ultimately endangers patient safety. Two feature sets—biometric and network flow metric—have been incorporated to enhance detection in healthcare systems. Another major challenge lies in the scarcity of publicly available balanced datasets for analyzing diverse IoMT attack patterns. To address this, the Auxiliary Classifier Generative Adversarial Network (ACGAN) was employed to generate synthetic samples that resemble minority class samples. ACGAN operates with two objectives: the discriminator differentiates between real and synthetic samples while also predicting the correct class labels. This dual functionality ensures that the discriminator learns detailed features for both tasks. Meanwhile, the generator produces high-quality samples that are classified as real by the discriminator and correctly labeled by the auxiliary classifier. The performance of this approach, evaluated using the IoMT dataset, consistently outperforms the existing baseline model across key metrics, including accuracy, precision, recall, F1-score, area under curve (AUC), and confusion matrix results.

A Comparative Study of Transformer-based Models for Hate-Speech Detection in English-Kiswahili Code-Switched Social Media Text

The transformer architecture, first introduced in 2017 by researchers at Google, has revolutionized natural language processing in various tasks, including text classification. This architecture formed the basis of future models such as those used in hate speech detection in code-switched text. In this research, we conduct a comparative study of transformer-based models for hate speech detection in English-Kiswahili code-switched text. First, the models were compared as feature extractors using a traditional classifier and then as end-to-end classifiers. The three multilingual transformer-based models compared include mBERT, mDistilBERT and XLM-RoBERTa, using SVM as the traditional classifier for the extracted features. The HateSpeech_Kenya dataset, sourced from Kaggle, was utilized in this study. As a feature extractor, mBERT’s hidden states trained the highest-performing SVM with an accuracy of 0.5461 and a macro f1 score of 0.40. Among the three models evaluated, XLM-RoBERTa achieved the highest accuracy of 0.6069 and a macro f1 score of 0.49 on a balanced dataset. In contrast, mBERT achieved the highest accuracy of 0.7820 and a macro f1 score of 0.53 on an imbalanced dataset. The comparative study establishes that using transformer-based models as end-to-end classifiers generally performs better than using them as feature extractors with traditional classifiers. This is because directly training the models allows them to learn more task-specific features. Furthermore, the varying performance across balanced and imbalanced datasets highlights the need for careful model selection based on the dataset characteristics and specific task requirements.

Comparative Analysis of Cross-Validation Techniques: LOOCV, K-folds Cross-Validation, and Repeated K-folds Cross-Validation in Machine Learning Models

Effective model evaluation is crucial for robust machine learning, and cross-validation techniques play a significant role. This study compares Repeated k-folds Cross Validation, k-folds Cross Validation, and Leave-One-Out Cross Validation (LOOCV) on imbalanced and balanced datasets across four models: Support Vector Machine (SVM), K-Nearest Neighbors (K-NN), Random Forest (RF), and Bagging, both with and without parameter tuning. On imbalanced data without parameter tuning, Repeated k-folds cross-validation demonstrated strong performance for SVM with a sensitivity of 0.541 and balanced accuracy of 0.764. K-folds Cross Validation showed a higher sensitivity of 0.784 for RF and a balanced accuracy of 0.884. In contrast, LOOCV achieved notable sensitivity for RF and Bagging at 0.787 and 0.784, respectively, but at the cost of lower precision and higher variance, as detailed in Table 1. When parameter tuning was applied to balanced data, the performance metrics improved. Sensitivity for SVM reached 0.893 with LOOCV and balanced accuracy for Bagging increased to 0.895. Stratified k-folds provided enhanced precision and F1-Score for SVM and RF. Notably, processing times varied significantly, with k-folds being the most efficient with SVM taking 21.480 seconds and Repeated k-folds showing higher computational demands where RF took approximately 1986.570 seconds in model processing, as shown in Table 4. This analysis underscores that while k-folds and repeated k-folds are generally efficient, LOOCV and balanced approaches offer enhanced accuracy for specific models but require greater computational resources. The choice of cross-validation technique should thus be tailored to the dataset characteristics and computational constraints to ensure optimal model evaluation.

Exploring Client Preferences for Psychological Counselors in a Chinese Online Health Community: Longitudinal Study.

Although online health communities are acknowledged for their role in bridging the supply-demand gap in mental health services, the client decision-making process in these environments remains underexplored. This study aimed to explore the impact of different signals presented on psychological counselors' home pages on clients' choices. Adopting signaling theory as the framework, this study classified information into online and offline signals and developed a theoretical model to examine client choice behaviors. We collected data from 487 psychological counselors in a leading Chinese online mental health community during March, June, September, and December 2023. Based on these data, we constructed a 4-period balanced panel dataset. A fixed effects model was used to analyze which signals influence clients' choices of psychological counselors. Regarding online signals, the service price (β=0.186, P<.001) and online reputation (β=0.489, P=.002) of psychological counselors positively influence clients' choices. Concerning offline signals, psychological counselors' practical experience (β=0.007, P<.001) is positively related to clients' choices. Moreover, the results indicate that the relationship between a counselor's prosocial behavior and clients' choices is not linear but rather exhibits an inverted U-shape. This study reveals that the varied information provided by psychological counselors has distinct impacts on clients' choices in online health communities. It broadens the application of signaling theory to online behaviors and emphasizes the importance of both online and offline signals. These insights offer strategic guidance for counselors and online platforms to better meet potential clients' needs by optimizing the information presented on psychological counselors' home pages.

Investigating the dynamic nexus between fiscal deficit and inflation: empirical evidence from BRICS economies

PurposeThe debate surrounding whether fiscal actions are responsible for inflation has been a central focus in macro-public finance. In this regard, the present study makes a novel attempt to assess the nexus between fiscal deficit and inflation in the presence of output growth, trade openness and money supply using a balanced panel dataset from Brazil, Russia, India, China, and South Africa (BRICS) economies.Design/methodology/approachIn conducting the empirical analysis, the study initially addresses the issues of cross-section dependency, heterogeneous slope coefficients and nonstationarity. The assessment of long-run cointegration was carried out by using novel third-generation cointegration tests introduced by Westerlund and Edgerton (2008) and Banerjee and Carrion-i-Silvestre (2017). The authors utilise the advanced and latest estimation technique, the cross-section augmented autoregressive-distributed lag (CS-ARDL) model, for long-run and short-run analyses. Finally, to investigate the causal relationship between fiscal deficit and inflation within a panel framework, this study employs the novel test (JKS) introduced by Juodis et al. (2021).FindingsUsing recent econometric tests, the study validates the existence of cross-sectional dependency and slope heterogeneity. The results of the third-generation cointegration test by Westerlund and Edgerton (2008) and Banerjee and Carrion-i-Silvestre (2017) show that the variables are cointegrated in the long run. The CS-ARDL model revealed a positive relationship between inflation and fiscal deficit. The implication of this finding suggests that the fiscal deficit plays an inflationary role in BRICS economies. Finally, the JKS causality test results found a bi-directional causal association between fiscal deficit and inflation.Practical implicationsThe findings suggest several significant policy recommendations. It is recommended that a well-executed fiscal consolidation strategy be adopted to attain sound fiscal health and lower inflation. A disciplined fiscal approach is not only vital for effective monetary policy but also essential for maintaining macroeconomic stability. Monetary authorities must establish credible practices to effectively manage the macroeconomic system, and policy stances should align with the specific needs of the economy. The bidirectional causality between fiscal deficit and inflation suggests that relying solely on fiscal measures is inadequate for managing inflationary pressures or stabilising fiscal balances. Thus, a more comprehensive approach is required, including measures such as sustaining economic growth, reducing import dependency, diversifying exports and ensuring exchange rate stability.Originality/valueNumerous empirical studies have explored the link between fiscal deficit and inflation. However, most research studies within BRICS economies have focused on individual countries rather than considering the group as a whole. This limited scope may fail to capture the unique characteristics and interactions within these economies. To the best of the author’s knowledge, the present study is the first attempt to examine the issue from a panel perspective across BRICS economies. The methodological novelty of the present study is that it represents the first attempt, at least within emerging market economies, to investigate the nexus between fiscal deficit and inflation using second- and third-generation econometric models. From a policy perspective, the authors highlight that BRICS economies must prioritise fiscal discipline through measures such as reducing unproductive expenditure and improving tax collection. Close coordination between fiscal and monetary authorities is essential to ensuring that monetary policy supports fiscal consolidation efforts while maintaining price stability.

Fault Diagnosis of Induction Motors under Limited Data for Across Loading by Residual VGG-Based Siamese Network

This study proposes an improved few-shot learning model of the Siamese network residual Visual Geometry Group (VGG). This model combined with time–frequency domain transformation techniques effectively enhances the performance of across-load fault diagnosis for induction motors with limited data conditions. The proposed residual VGG-based Siamese network consists of two primary components: the feature extraction network, which is the residual VGG, and the merged similarity layer. First, the residual VGG architecture utilizes residual learning to boost learning efficiency and mitigate the degradation problem typically associated with deep neural networks. The employment of smaller convolutional kernels substantially reduces the number of model parameters, expedites model convergence, and curtails overfitting. Second, the merged similarity layer incorporates multiple distance metrics for similarity measurement to enhance classification performance. For cross-domain fault diagnosis in induction motors, we developed experimental models representing four common types of faults. We measured the vibration signals from both healthy and faulty models under varying loads. We then applied the proposed model to evaluate and compare its effectiveness in cross-domain fault diagnosis against conventional AI models. Experimental results indicate that when the imbalance ratio reached 20:1, the average accuracy of the proposed residual VGG-based Siamese network for fault diagnosis across different loads was 98%, closely matching the accuracy of balanced and sufficient datasets, and significantly surpassing the diagnostic performance of other models.

Create a Realistic IoT Dataset Using Conditional Generative Adversarial Network

The increased use of Internet of Things (IoT) devices has led to greater threats to privacy and security. This has created a need for more effective cybersecurity applications. However, the effectiveness of these systems is often limited by the lack of comprehensive and balanced datasets. This research contributes to IoT security by tackling the challenges in dataset generation and providing a valuable resource for IoT security research. Our method involves creating a testbed, building the `Joint Dataset’, and developing an innovative tool. The tool consists of two modules: an Exploratory Data Analysis (EDA) module, and a Generator module. The Generator module uses a Conditional Generative Adversarial Network (CGAN) to address data imbalance and generate high-quality synthetic data that accurately represent real-world network traffic. To showcase the effectiveness of the tool, the proportion of imbalance reduction in the generated dataset was computed and benchmarked to the BOT-IOT dataset. The results demonstrated the robustness of synthetic data generation in creating balanced datasets.

Digital brains, green gains: Artificial intelligence's path to sustainable transformation

The environmental impacts of artificial intelligence on a global scale remain underexplored. This study utilizes a balanced panel dataset to examine artificial intelligence's complex role in enhancing global green productivity between 2008 and 2019. The findings indicate that artificial intelligence robustly boosts green productivity, even after correcting for potential endogeneity using the legal system's origin as an instrument. A detailed mediation analysis underscores that artificial intelligence indirectly promotes green productivity by increasing renewable energy use, attracting skilled labor, and dampening stock market performance. Additional analysis confirms that financial development generally amplifies artificial intelligence's favorable effects on green productivity. However, the combined impact of financial institution access and artificial intelligence on green productivity initially appears hostile, an effect that can be reversed when financial access exceeds a certain threshold. These results offer valuable insights into the interconnection between artificial intelligence and the global shift towards greener practices.

B-111 Advancing Precision Medicine in Multiple Myeloma: Addressing Demographic Variabilities and Imbalanced Data in the NIH All of Us Research Program Cohort

Abstract Background The diagnosis of multiple myeloma (MM), a complex hematologic malignancy, presents unique challenges in prognostication and treatment planning, exacerbated by disparities across demographic groups and imbalanced datasets. The heterogeneity observed in MM's clinical manifestations and outcomes necessitates a nuanced approach to patient care, emphasizing the need for models that accurately predict long-term health outcomes across the diverse spectrum of MM participants. Given the variable response to treatment observed among different demographic cohorts, achieving precision in treatment strategies and ensuring equitable healthcare provision is paramount. Objective The objective of this study was to develop and assess predictive models tailored for various demographic groups of participants diagnosed with MM in the All of Us Research Program cohort, which is enrolling 1 million diverse participants who share multiple sources of data. Methods The Synthetic Minority Over-Sampling Technique (SMOTE) was used to rectify the prevalent imbalances within datasets, thereby significantly enhancing the predictive accuracy of patient outcomes over a 1-year post-diagnosis period. By addressing these imbalances, we provide a cornerstone for the development of more personalized and effective treatment paradigms. Employing a comprehensive dataset extracted from electronic health records (EHRs), we segmented participants based on key demographic variables such as age, gender, race, and socioeconomic status. This segmentation included detailed analyses of MM diagnosis, treatment regimens, and subsequent patient responses. To tackle the issue of dataset imbalance, two sets of predictive models were meticulously crafted for each demographic segment: one utilized the original dataset distributions and the other leveraged datasets balanced with SMOTE. A suite of performance metrics, including accuracy, precision, recall, Area Under the Curve (AUC), and F1 score, were deployed to evaluate and compare the efficacy of these models. Models trained on SMOTE-balanced datasets demonstrated marked improvements in prediction accuracy, from a baseline of 75% to over 90%. Precision rates climbed to 88%, and recall rates reached 90%, with an AUC of 0.95 and an F1 score of 0.89, highlighting the models' enhanced predictive capabilities. Results The application of SMOTE to balance datasets resulted in significant predictive accuracy improvements for minority groups within the MM patient population. This was especially evident in demographic cohorts historically underrepresented in clinical research, showcasing the potential of SMOTE in creating more equitable treatment outcomes. Conclusions Our findings underscore the potential of employing tailored predictive models in refining prognosis and treatment planning for multiple myeloma, particularly by addressing data imbalances. The integration of SMOTE into our methodology validates its efficacy in enhancing prediction accuracy, setting a new standard for oncological research. Implications Advocating for a more sophisticated and inclusive approach to data handling, our research contributes to a paradigm shift in oncological research, especially within hematological malignancies like MM. This study is poised to revolutionize predictive modeling in oncology, fostering more informed clinical decision-making and advancing the goals of precision medicine.

Enhancing aspect category detection in imbalanced online reviews: An integrated approach using Select-SMOTE and LightGBM

Aspect category detection (ACD) is a pivotal subtask within the field of sentiment analysis in natural language processing, aiming to identify implicit aspect category information in online review texts. In real-world scenarios of online review category detection tasks, data imbalance often arises, leading to skewed distributions among distinct review categories. This phenomenon poses substantial challenges for accurately recognizing minority categories through modeling. To address this, we propose a method for detecting imbalanced aspect categories by combining the selective synthetic over-sampling (Select-SMOTE) algorithm with the light gradient boosting machine (LightGBM). Our approach commences with text data representation through features, followed by a strategy involving joint sample partitioning and boundary optimization within the feature space to generate minority class samples. This partitioning strategy aligns generated data more closely with the original distribution, while the boundary optimization module enhances classification performance by eliminating samples near boundaries. Subsequently, the balanced dataset is input to the LGB model, enabling the extraction of aspect category information through parameter optimization and class weight assignment. Finally, our method is evaluated using the SemEval and SentiHood datasets and compared with prevailing sampling methods and classification models. Empirical results manifestly demonstrate the method’s superiority across diverse metrics, reflecting robustness and effective mitigation of imbalanced data challenges in ACD.

Adaptive capacity to climate change: Does energy aid matter?

The global call for augmented financial support to satisfy the growing requirements for adaptation funding, essential for enabling vulnerable groups to withstand the ramifications of climate change, has been resonant. Channeling energy assistance to developing countries may serve to ameliorate this shortfall. Through an empirical analysis using a balanced panel dataset comprising 64 countries from 2002 to 2020, the research investigates the role of energy assistance in enhancing the adaptive capacities of developing nations. Our analysis identifies critical areas with pronounced deficiencies in adaptive capacity, primarily situated in sub-Saharan Africa, South Asia, Central Asia, and western South America. The study reveals that energy aid, notably in the forms of non-renewable, policy-related, and distribution assistance, significantly elevates adaptive capabilities. Although this assistance yields significant benefits for countries across various income levels, the impact is relatively greater for higher-income nations. Additionally, energy aid indirectly bolsters adaptive capacity by stimulating innovation, and an improvement in the quality of governance aids in enhancing the effectiveness of energy assistance implementation, especially for lower-income countries. The study concludes by offering nuanced policy insights aimed at donors and recipients alike, with the goal of augmenting the efficacy of aid to improve climate change adaptation.

ACGAN: adaptive conditional generative adversarial network architecture predicting skin lesion using collaboration of transfer learning models

Skin cancer has become a serious disease which has the potential to scale up if it is not identified earlier. It is imperative to detect and give treatment to skin cancer promptly. Diagnosing skin cancer manually takes a lot of time and it is costly, and the probability of false diagnosis has increased due to the outstanding resemblances among various skin lesions. Enhancing the classification of multi-class lesions of skin needs the development of investigative systems which should be automated. Data augmentation with GANs and Adaptive Conditional Generative Adversarial Network strategies improves performance. The performance is tested using balanced and unbalanced datasets. Using a proper process of augmentation of data, the suggested system attains a 94% accuracy for the VGG16, 93% for the ResNet50 and 94.25% for ResNet101. The process of collaboration of all such methods improves accuracy further to 95%. In summary, the novelty of the work lies in its holistic approach to automated skin lesion classification, incorporating advanced deep learning models, novel data augmentation techniques and comprehensive performance evaluation on real-world datasets. These contributions collectively advance the field of computer-aided diagnosis for the detection of skin cancer and treatment.

Performance Analysis of Garbage Classification Using Balanced and Unbalanced Dataset with EfficientNetV2B1 Architecture

Performance Analysis of Garbage Classification Using Balanced and Unbalanced Dataset with EfficientNetV2B1 Architecture

EFFECTS OF MICROCREDIT PROGRAMS ON HOUSEHOLD FINANCIAL PORTFOLIOS: EVIDENCE FROM RURAL VIETNAM

Although microcredit is generally regarded as an excellent instrument for assisting borrowers in improving their well-being, its contribution to improving family financial status is often overlooked. This study addresses this issue by examining the effects of microcredit programs in Vietnam on household financial portfolios. We utilize a balanced panel dataset of 1157 identical households from two Vietnamese provinces collected in 2016, 2017, and 2022. The estimated results are calculated using the multivariate Tobit method and the instrumental variable regression method. The findings prove that microcredit increases households' lending, savings, and investments. The results are consistent with the conceptual framework and robust to different econometric models and samples. The study suggests some policy implications based on its findings.

Artificial intelligence abnormal driving behavior detection for mitigating traffic accidents

Annually, a tragic toll of 1.3 million lives is lost on roads across the globe, with tens of millions more suffering injuries or disabilities. The necessity for precise detection of abnormal driving behavior is paramount in reducing traffic accidents. This paper aims to bridge the gap between normal and abnormal driving patterns, offering near-flawless detection capabilities. This paper presents a novel AI tachograph prototype, the first of its kind, that can classify driving behavior into normal and abnormal in real time with an impressive accuracy of 99.99 %. This high level of accuracy is achieved by using a bias-reduction method. The bias-reduction method focuses on minimizing biases in the dataset, such as surrounding situations, location, driver information, and car types. This approach significantly enhances the prediction accuracy of existing machine learning algorithms. The dataset used for this research is quite extensive, consisting of anomaly data collected from 10,181 commercial vehicles and 12,530 drivers in just 0.1 s. This rich dataset is crucial for building a reliable model. The effectiveness of the proposed method was validated using 10-fold cross validation on 480 k to 540 k instances with 36 determinants. The results clearly demonstrated that reducing bias leads to higher prediction accuracy. The paper also plans to compare the prediction accuracy of balanced and imbalanced datasets. The findings from this research have broader implications as the proposed method can be applied generally to machine learning to improve prediction accuracy.

Multi-attribute balanced dataset generation framework AutoSyn and KinFace Channel-Spatial Feature Extractor for kinship recognition

In the field of kinship verification, facial recognition technology is becoming increasingly vital due to privacy concerns, ethical disputes, and the high costs associated with DNA testing. We have developed a novel method, the AutoSyn framework, to synthesize facial images and enhance kinship image datasets, effectively addressing the challenges of scale and quality in existing datasets. By employing a strategy that mixes ages and genders in the synthesized images, we minimize the impact of these factors on kinship recognition tasks. We have enhanced the original KinFaceW-I dataset by integrating ten distinct styles, including diverse combinations of gender, ethnicity, and age. This enrichment significantly improves both the quality and quantity of the images. Furthermore, this paper introduces an efficient feature extractor for kinship tasks, KinFace-CSFE, within a siamese neural network framework. This model not only utilizes meticulously designed channel feature extraction but also incorporates mixed kernel size spatial attention mechanisms to better focus on local features. We have also integrated YOCO data augmentation techniques to simulate complex imaging conditions, enhancing the model’s robustness and accuracy. The effectiveness of these innovations has been validated through experiments on the KinFaceW-I, KinFaceW-II, and synthesized Syn-KinFaceW-I datasets, achieving accuracy rates of 82.7%, 94.1%, and 83.2% respectively. These results significantly surpass both traditional models and current advanced models.

An adaptive binary classifier for highly imbalanced datasets on the Edge

Edge machine learning brings intelligence to low-power devices at the periphery of a network. By running machine learning algorithms on the Edge, classification can be performed faster without the need to transmit large data volumes across a network. However, on-device training is often not feasible since Edge devices have limited computing and storage resources. Improved, Scalable, Efficient, and Fast classifieR (iSEFR) is a classifier that performs both training and testing on low-power devices using linearly separable balanced datasets. The novelty of this work is the improvement of the iSEFR accuracy by fine-tuning the algorithm with datasets having an uneven class distribution. Three adaptive linear function transformation techniques were proposed to improve the decision threshold which is in the form of a linear function. Experiments using stratified sampling with 5-fold cross-validation demonstrate that one of the proposed techniques significantly improved F1-score, Recall and Matthews Correlation Coefficient (MCC) by an average of 23%, 35% and 21% compared to iSEFR. Further evaluation of this technique in a Fog environment using highly imbalanced datasets such as credit card fraud, network intrusion and diabetic retinopathy also showed a significant increase of 38%, 44% and 30% in F1-score, Recall and MCC with a Precision of 97%. The adaptive binary classifier maintained the time complexity of iSEFR without altering the class imbalance.