Class Distribution Research Articles

A Framework for Diabetic Retinopathy Detection using Transfer Learning and Data Fusion

Diabetic retinopathy stands as a significant concern for individuals managing diabetes. It is a severe eye condition that targets the delicate blood vessels within the retina. As it advances, it can inflict severe vision impairment or complete blindness in extreme cases. Regular eye examinations are vital for individuals with diabetes to detect abnormalities early. Detection of diabetic retinopathy is challenging and a time-consuming process, but deep learning and transfer learning techniques offer vital support by automating the process, providing accurate predictions, and simplifying diagnostic procedures for healthcare professionals. This study introduces a multi-classification framework for grading diabetic retinopathy into five classes using Transfer Learning and data fusion. The objective is to develop a robust, automated model for diabetic retinopathy detection to enhance the diagnostic process for healthcare professionals. We fused two distinct datasets, APTOS and IDRiD, which resulted in a total of 4178 fundus images. The merged dataset underwent preprocessing to enhance image quality and to remove unwanted regions, noise and artifacts from the fundus images. The pre-processed dataset is then resized and a balancing technique called SMOTE is applied to it due to uneven class distribution present among classes. To increase diversity and size of the dataset, data augmentation techniques including flipping, brightness adjustment and contrast adjustment are applied. The dataset is split into 80:10:10 ratios for training, validation, and testing. Two pre-trained models, EfficientNetB5 and DenseNet121, are fine-tuned and training parameters like batch size, number of epochs, learning rate etc. are adjusted. The results demonstrate the highest test accuracy of 96.06% is achieved by using EfficientNetB5 model followed by 91.40% test accuracy using DenseNet121 model. The performance of our best model i.e. EfficientNetB5, is compared with several state-of-the-art approaches, including DenseNet-169, Hybrid models and ResNet-50 where our model outperformed these methodologies in terms of test accuracy.

ChromTR: chromosome detection in raw metaphase cell images via deformable transformers.

Chromosome karyotyping is a critical way to diagnose various hematological malignancies and genetic diseases, of which chromosome detection in raw metaphase cell images is the most critical and challenging step. In this work, focusing on the joint optimization of chromosome localization and classification, we propose ChromTR to accurately detect and classify 24 classes of chromosomes in raw metaphase cell images. ChromTR incorporates semantic feature learning and class distribution learning into a unified DETR-based detection framework. Specifically, we first propose a Semantic Feature Learning Network (SFLN) for semantic feature extraction and chromosome foreground region segmentation with object-wise supervision. Next, we construct a Semantic-Aware Transformer (SAT) with two parallel encoders and a Semantic-Aware decoder to integrate global visual and semantic features. To provide a prediction with a precise chromosome number and category distribution, a Category Distribution Reasoning Module (CDRM) is built for foreground-background objects and chromosome class distribution reasoning. We evaluate ChromTR on 1404 newly collected R-band metaphase images and the public G-band dataset AutoKary2022. Our proposed ChromTR outperforms all previous chromosome detection methods with an average precision of 92.56% in R-band chromosome detection, surpassing the baseline method by 3.02%. In a clinical test, ChromTR is also confident in tackling normal and numerically abnormal karyotypes. When extended to the chromosome enumeration task, ChromTR also demonstrates state-of-the-art performances on R-band and G-band two metaphase image datasets. Given these superior performances to other methods, our proposed method has been applied to assist clinical karyotype diagnosis.

A self-supervised deep-driven model for automatic weather classification from remote sensing images

ABSTRACT The increasing use of Remote Sensing (RS) technology has resulted in the availability of large volumes of satellite imaging data, necessitating efficient, and scalable solutions for rapid analysis and classification in various interdisciplinary fields. This research focuses on addressing these challenges by examining different training and optimization techniques to achieve the highest classification accuracy on both labelled and unlabelled data from the Large-Scale Cloud Photos Dataset for Meteorology Research (LSCIDMR), which contains 10 high-resolution classes. A key aspect of this study is the implementation of a unique deep learning model based on the ResNet101 architecture, designed to handle the inter-class similarity problem through architectural improvements and imbalanced class distributions using data augmentation techniques. Furthermore, this model integrates a novel approach that combines self-learning with semi-supervised learning to process large amounts of unlabelled data. An iterative self-learning process is employed for continuous refinement, using pseudo-labels generated from the unlabelled data. To ensure higher quality and relevance, confidence-based selection of pseudo-labels is applied. Our model demonstrates superior performance in classification compared to recently reported deep learning algorithms, offering a robust solution for the automated categorization of satellite images. In this study, our ResNet101 model achieved a mean accuracy of 0.917 (±0.02), a mean precision of 0.917 (±0.01), with a mean F1 score of 0.9300 (±0.015) over five folds.

Tree populations show low regeneration of valued species in West Africa

Tree populations have declined substantially in West Africa in recent decades, raising concerns since trees provide numerous ecosystem goods and services. Regional information on the population status of tree species could guide more effective conservation and regeneration of natural vegetation. Here, we report results of the first regional analysis of tree population structure across the Sahel and Sudan zones, a meta-study of vegetation inventories, including 23,586 individual trees sampled across nine countries. We evaluated current status and forecast future trends of 16 species and one genus of trees of ecological and socio-economic importance. Size class distribution (SCD) reflects the population structure of an individual species and can provide early warning of composition change and population decline. SCD is analysed widely at a local scale, but analysis at a regional scale is needed to detect widespread population changes. Many native species lacked trees in the smaller size classes, implying unsustainable populations and future decline. Some species show sound regeneration at the regional scale, but high variation among sites. Eight species, including Adansonia digitata and Afzelia africana, show regional declines in regeneration and risks of future extirpations. Four of these severely lack regeneration. Protected areas show higher tree regeneration, but protected status did not assure good regeneration. Our results identify priority tree species across West Africa, indicate a more urgent need for conservation and regeneration of native tree species, and highlight the benefit of effective conservation. More widespread protection could increase tree populations, conserving biodiversity, and ecosystem services essential for people's livelihoods.

Triplet Adaptation Framework for Robust Semi-Supervised Learning.

Semi-supervised learning (SSL) suffers from severe performance degradation when labeled and unlabeled data come from inconsistent and imbalanced distribution. Nonetheless, there is a lack of theoretical guidance regarding a remedy for this issue. To bridge the gap between theoretical insights and practical solutions, we embark to an analysis of generalization bound of classic SSL algorithms. This analysis reveals that distribution inconsistency between unlabeled and labeled data can cause a significant generalization error bound. Motivated by this theoretical insight, we present a Triplet Adaptation Framework (TAF) to reduce the distribution divergence and improve the generalization of SSL models. TAF comprises three adapters: Balanced Residual Adapter, aiming to map the class distribution of labeled and unlabeled data to a uniform distribution for reducing class distribution divergence; Representation Adapter, aiming to map the representation distribution of unlabeled data to labeled one for reducing representation distribution divergence; and Pseudo-Label Adapter, aiming to align the predicted pseudo-labels with the class distribution of unlabeled data, thereby preventing erroneous pseudo-labels from exacerbating representation divergence. These three adapters collaborate synergistically to reduce the generalization bound, ultimately achieving a more robust and generalizable SSL model. Extensive experiments across various robust SSL scenarios validate the efficacy of our method.

Mine Your Own Anatomy: Revisiting Medical Image Segmentation With Extremely Limited Labels.

Recent studies on contrastive learning have achieved remarkable performance solely by leveraging few labels in the context of medical image segmentation. Existing methods mainly focus on instance discrimination and invariant mapping (i.e., pulling positive samples closer and negative samples apart in the feature space). However, they face three common pitfalls: (1) tailness: medical image data usually follows an implicit long-tail class distribution. Blindly leveraging all pixels in training hence can lead to the data imbalance issues, and cause deteriorated performance; (2) consistency: it remains unclear whether a segmentation model has learned meaningful and yet consistent anatomical features due to the intra-class variations between different anatomical features; and (3) diversity: the intra-slice correlations within the entire dataset have received significantly less attention. This motivates us to seek a principled approach for strategically making use of the dataset itself to discover similar yet distinct samples from different anatomical views. In this paper, we introduce a novel semi-supervised 2D medical image segmentation framework termed Mine yOur owNAnatomy (MONA), and make three contributions. First, prior work argues that every pixel equally matters to the model training; we observe empirically that this alone is unlikely to define meaningful anatomical features, mainly due to lacking the supervision signal. We show two simple solutions towards learning invariances-through the use of stronger data augmentations and nearest neighbors. Second, we construct a set of objectives that encourage the model to be capable of decomposing medical images into a collection of anatomical features in an unsupervised manner. Lastly, we both empirically and theoretically, demonstrate the efficacy of our MONA on three benchmark datasets with different labeled settings, achieving new state-of-the-art under different labeled semi-supervised settings. MONA makes minimal assumptions on domain expertise, and hence constitutes a practical and versatile solution in medical image analysis. We provide the PyTorch-like pseudo-code in supplementary.

Geological remote sensing interpretation via a local-to-global sensitive feature fusion network

Interpreting surface geological elements (such as rocks, minerals, soils, and water bodies) is the main task of geological survey, which plays a crucial role in geological environment remote sensing (GERS). However, the characteristics of geological elements, including high variabilities, various morphology, complicated boundaries and imbalanced class distribution, make it still a challenge for deep learning methods to interpret GERS images. Considering the correlations of geological elements as the regionalized variables in geostatistics, the sensitive features of GERS interpretation mainly include three aspects: tonal, textural and structural characteristics within a singular-class elements, spatial and spectral correlations of adjacent elements, and their global tectonic or spatial distribution. Thus, to simulate the manual interpretation process of geologists from local to global and promote GERS interpretation performance, we propose a local-to-global multi-scale feature fusion network (LGMSFNet). A geological object context represents the intra-class semantic dependencies of pixel sets with the same class. And a local feature aggregation module models the channel and spatial association. Then discriminative features are integrated by a global feature fusion module. For the model optimization, we focus on hard examples during the training process to achieve the balanced optimization of various categories. Two research areas that include large-scale rocks, soils and water exposed on the surface are selected. Massive experiments demonstrate the superiority of the LGMSFNet in GERS interpretation.

A probabilistic approach driven credit card anomaly detection with CBLOF and isolation forest models

Businesses throughout the world are starting to accept more and more electronic payment alternatives for in-person and online purchases. Unusual behaviour such as internet fraud and payment defaults, which may lead to significant monetary losses, have become increasingly widespread as credit card use in online buying has expanded. To find a solution to this problem, researchers looked into a number of different machine learning classifiers that may identify irregularities in the data pertaining to credit card transactions. Overlapping class samples and an uneven distribution of classes make it hard to find outliers in this data. General learning algorithms may favour samples from the majority class, and the detection rate of anomalies in samples from minority classes is thus low. Our proposed Credit Card Outlier Detection (CCOD) model incorporates many machine learning algorithms to enhance the detection rates of credit card anomalies. Utilising the stratified sampling technique and the k-fold cross-validation process, we address data imbalance and overfitting. Working with optimized selected features rather than all features reduces the danger of overfitting and improves model performance. It has been noted that the findings of Cluster-Based Local Outlier Factor (CBLOF) classifier on European Credit Card Dataset and Isolation Forest classifier on German credit card dataset were superior to those of the other classifiers with Mathew correlation coefficient value of 0.95 and 0.97 respectively. The CBLOF, an advanced outlier detection method that evaluates the local density of data points within clusters. By comparing the density of a point to the densities of its surrounding clusters, CBLOF identifies outliers based on their deviation from normal cluster behaviour. In general, this CCOD model shows that it is better at dealing with issues like uneven class distribution, overfitting, and classes that overlap.

Forecasting energy poverty using different machine learning techniques for Missouri

Energy poverty in Missouri was analyzed using the four quadrant approach using both state and county level data sets for two separate definitions of the grid. Predictions used machine learning techniques including decision trees, random forest, extreme gradient boosting and support vector machines. It was determined that the extreme gradient boosting performed the best when compared to all the other models after the hyperparameters were tuned. The F1 scores for the county level data sets were higher than for the state levels thus indicating greater predictability for the National Oceanic and Atmospheric Administration (NOAA) climatological regional runs. For the county level data, the F1 score was the highest for region 1, which coincided with one of the highest expenditure risk values whilst regions 2-4 were the lowest scoring area. In grid 2, the largest class distribution changed from grid 1’s expenditure risk to the no risk category. This grid had more variability in terms of the double risk class when compared to grid 1 and, as such, its predictability in terms of its F1 scores was reduced. There were similarities in the ranking of the prediction scores for the regions for both grids as regions 1 and 6 incurred the largest F1 values. Thus energy poverty can be classified and predicted for Missouri, which in turn may aid policy makers via quantitative regional risk analysis. This data-driven informed policy making can lead to the development and implementation of laws and social programs to help ameliorate energy poverty.

Spatial distribution and transport characteristics of debris flow sediment using high resolution UAV images in the Ohya debris flow fan

Debris flows are significant geomorphic processes that can pose hazards to human settlements and infrastructure, making the understanding of their sediment transfer characteristics and impacts on the surrounding landscape crucial for effective land management and hazard mitigation strategies. This study investigates the distribution patterns of sandstone, vegetation, and shale classes within the Ohya debris flow fan, aiming to gain insights into the sediment characteristics, grain size variations, and vegetation cover trends along the flow path. Using a combination of field surveys and image processing we analyzed sediment characteristics and grain size variations a three study zones: upstream, transfer, and downstream. Statistical parameters such as mean, mode, skewness, and kurtosis were employed to quantify the grain size distribution and identify variations across the study zones. Results reveal distinct spatial variations in the distribution of sandstone, vegetation, and shale classes along the flow path. We observed a decrease in sandstone and vegetation cover from the upstream to the downstream sections, while the area covered by the shale class exhibited an increasing trend in the downstream direction. These findings suggest that debris flows selectively remove smaller particles downstream. The distribution of particle sizes within the sandstone class also exhibited variations, with a higher proportion of smaller pebbles downstream due to more effective removal of finer particles in the upstream zone. This study highlights the complex interplay between debris flow dynamics, sediment transport, and vegetation colonization. The findings contribute to the scientific knowledge of debris flow processes and can be used to inform land management decisions and mitigate the hazards associated with debris flows in similar environments.

Woody Species Composition, Structure, and Diversity of Parkland and Coffee-Based Agroforestry Systems, Habro District, Eastern Ethiopia

Abstract Agroforestry practices in eastern Ethiopia are renowned for their multifunctional landscapes, contributing significantly to biodiversity conservation and enhancement. However, there is a lack of comprehensive studies on the diversity and structure of woody species within coffee-based and parkland agroforestry systems documented in a single study. This research aimed to evaluate the composition, diversity, and structure of woody species in these agroforestry systems. Conducted in the Habro district of eastern Ethiopia, the study involved randomly selecting sixteen plots for each agroforestry system. Inventory assessments of woody species were carried out using 40×40 and 20 m*20-meter plots for parkland and coffee-based systems respectively, and 5×5-meter plots for coffee shrubs. For trees with a diameter at breast height (DBH) of ≥2.5 cm, measurements of tree heights were taken. The study computed basal area, stem density, diameter, and height class distribution to characterize the structure of woody communities, and species diversity was also assessed. The findings revealed a total of 57 woody species across 31 families. Specifically, 38 woody species were recorded in parkland agroforestry, while 43 species were found in coffee-based agroforestry systems. Significant differences were observed in species diversity indices and structural parameters between the two agroforestry systems. The Shannon diversity index and richness were higher in coffee-based agroforestry compared to parkland systems. Additionally, the density and basal area of woody species were greater in coffee-based systems than in parkland agroforestry. Overall, both agroforestry systems were found to conserve a significant number of woody species, highlighting their potential to contribute to biodiversity conservation and informing future agroforestry management strategies in national programs.

Overlap to equilibrium: Oversampling imbalanced datasets using overlapping degree

Imbalanced and overlapping class distributions present several challenges, including poor generalization, misleading accuracy, and inflated importance of the majority class, which further complicate the classification task. To tackle this, we introduce a new novel oversampling method called GOS that generates samples from positive overlapping samples for imbalanced and overlapping data which improves the classification performance. Firstly, In GOS, a novel concept termed overlapping degree is introduced utilizing both local and global information from positive and negative samples. Secondly, it measures how much a positive sample contributes to the overlapping region and helps to identify positively overlapping samples. Lastly, the identified positive overlapping samples are transformed to generate new positive samples with a transformation matrix derived from the distribution information of all positive samples. We compare GOS with 14 commonly used under-sampling, oversampling, and advanced oversampling methods on 15 publicly available real imbalanced datasets with sample sizes varying from 178 to 2000 having an imbalance ratio varying from 2.02 to 41.4. The experimental results show that GOS outperforms these baselines achieving average improvements of 3.2 % in accuracy, 2.5 % in G-mean, 4.5 % in F1-score, and 5.2 % in AUC.

Robust diabetic prediction using ensemble machine learning models with synthetic minority over-sampling technique

This paper addresses the pressing issue of diabetes, which is a widespread condition affecting a huge population worldwide. As cells become less responsive to insulin or fail to produce it adequately, blood sugar levels rise. This has the potential to cause severe health complications including kidney disease, vision impairment and heart conditions. Early diagnosis is paramount in mitigating the risk and severity of diabetes-related complications. To tackle this, we proposed a robust framework for diabetes prediction using Synthetic Minority Over-sampling Technique (SMOTE) with ensemble machine learning techniques. Our approach incorporates strategies such as imputation of missing values, outlier rejection, feature selection using correlation analysis and class distribution balancing using SMOTE. The extensive experimentation shows that the proposed combination of AdaBoost and XGBoost shows exceptional performance, with an impressive AUC of 0.968+/-0.015. This outperforms not only alternative methodologies presented in our study but also surpasses current state-of-the-art results. We anticipate that our model will significantly improve diabetes prediction, offering a promising avenue for improved healthcare outcomes in diabetes management.

Regeneration and Diameter-Height Distribution Under Irregular Shelterwood System: A Case Study from Banke District

This study was aimed to assess the regeneration composition and status and species diversity of blocks of compartment C4S6 with 8 sub-compartment C4S1, C4S2, C4S3, C4S4, C4S5, C4S6, C4S7, And C4S8 under Irregular Shelterwood System (ISS) based scientific forest management in Samshergunj-Mathebas block forest, Banke district, Nepal. The systematic random sampling method was used for allocating circular sample plots (of radius 1.78 m for seedlings and 2.82 m for sapling) for collecting data from the field. Altogether 40 sample plots were taken to collect the required data from the blocks forest. The total regeneration per hectare in the block is 76870. In the blocks forest the total number of seedlings and saplings per hectare was 65150 and 11720 respectively. The IVI value showed that both seedling and sapling of <i>Shorea robusta</i> was dominant species and <i>Terminalia alata </i>was 1st codominant species in the blocks. Vegetation study is crucial for balancing the ecosystem as well as for biophysical environment. Forest inventories helps to determine the growing stock as well as annual increment in order to balance between harvesting and re-growth. In my study area there is abundant regeneration in the forest but their effective assessment is lacking and research on it is insufficient. Due to the lack of sustainable management, open grazing, illegal harvesting of wood, and forest fire, Productivity of forest is decreasing day by day. And at present, there can be seen Asna (<i>Terminalia alata</i>), and other Sal associates like Botdhairo instead of Sal. Graduation of seedling into sapling with time is key necessity in maintaining sustainable forest regeneration. Thus, this study was conducted for assessing regeneration composition and diameter and height class distribution after implementation of irregular Shelterwood system (ISS) in western terai. This study will be useful for policy makers working in conservation biology.

Effects of sacubitril/valsartan according to background beta-blocker therapy in patients with heart failure and reduced ejection fraction: Insights from PARADIGM-HF.

Beta-blockers may inhibit neprilysin activity and conversely, neprilysin inhibition may have a sympatho-inhibitory action. Consequently, sacubitril/valsartan may have a greater effect in patients not receiving a beta-blocker compared to those treated with a beta-blocker. We examined the effect of sacubitril/valsartan compared to enalapril on outcomes according to background beta-blocker treatment in the 8399 patients with heart failure with reduced ejection fraction enrolled in PARADIGM-HF. The primary outcome was time to first heart failure hospitalization or cardiovascular death. Compared to the 7811 patients taking a beta-blocker, the 588 patients not receiving a beta-blocker were older, more frequently female, but had a similar mean left ventricular ejection fraction and New York Heart Association class distribution, with little difference in N-terminal pro-B-type natriuretic peptide. Patients not taking beta-blockers had a higher rate of the primary endpoint than those taking beta-blockers. The benefit of sacubitril/valsartan on the primary endpoint was evident in both the no beta-blocker subgroup (hazard ratio [HR] 0.61, 95% confidence interval [CI] 0.45-0.82) and the beta-blocker subgroup (HR 0.82, 95% CI 0.75-0.90; p-interaction = 0.06). The respective HRs for cardiovascular death were 0.47 (95% CI 0.32-0.69) versus 0.84 (95% CI 0.75-0.95; p-interaction <0.01) and for HF hospitalization 0.76 (95% CI 0.51-1.12) versus 0.80 (95% CI 0.71-0.90; p-interaction = 0.73). For all-cause death, the HR in the no beta-blocker group was 0.50 (95% CI 0.36-0.71) compared to 0.89 (95% CI 0.80-0.99) in the beta-blocker group (p-interaction <0.01). Safety outcomes related to sacubitril/valsartan versus enalapril did not differ according to background beta-blocker use. Sacubitril/valsartan may be more effective than enalapril in reducing the risk of death in patients not treated with a beta-blocker compared to those treated with a beta-blocker, but is effective regardless of beta-blocker use. ClinicalTrials.gov NCT01035255.

Exploring the Vegetation Characteristics of Karst Landscapes: A Study of Community Forest in Tubokarto Village, Wonogiri, Indonesia

Karst landscapes are characterized by unique relief and drainage patterns, which influence the types of vegetation that can inhabit these environments. Vegetation in karst ecosystems plays a crucial role in groundwater filtration. Plant species diversity within these communities contributes to structural variations across different habitats. This study evaluates the potential of vegetation in community forests in Tubokarto Village, Wonogiri Regency, by analyzing species composition and vegetation. Using a purposive sampling method, we selected areas with polyculture planting patterns and agroforestry, applying a sampling intensity of 2.2% with 20 × 100 m strip plots. The results showed that the species composition at the location consisted of 18 species in 7 families. The stand density for trees is 221 ind ha–1, poles 436 ind ha–1, saplings 1,144 ind ha–1, and seedlings 2,650 ind ha–1. Tectona grandis emerged as the dominant species, with the highest Importance Value Index (IVI) values: 142.79 at the tree level, 212.48 for poles, 197.84 for saplings, and 122.29 for seedlings. Species diversity indices were 1.53 for trees, 0.97 for poles, 1.36 for saplings, and 1.21 for seedlings. The species richness index showed values of 2.79 for trees, 1.49 for poles, 1.21 for saplings, and 1.80 for seedlings, while the species evenness index ranged from 0.43 to 0.76. The vegetation structure of Tubokarto Village spans strata B and C, with tree diameter class distribution resembling that of natural forests. Environmental factors, including climate and soil composition, are likely influencing species growth patterns.

Woody Species Population Structure and Regeneration Status of Hereje Natural Forest, Southwest Ethiopia

Determining the population structure and regeneration status of woody species in Hereje Natural Forest, Ethiopia, is essential for identifying potential management interventions to mitigate the adverse effects of deforestation and forest degradation. Three parallel transect lines (every 100 m) were established along altitudinal gradients, and 45 plots of 20 m x 20 m were used to collect vegetation data. Living woody species diameter breast height (DBH) was measured in each plot for those with height ≥2 m and DBH ≥ 2.5 cm at about 1.3 m from the ground. For this study seedlings and saplings inventory was made using five subplots of 2 m X 2m, four on each corner of the main plots and one in the center, In this study we have used appropriate formulas to analyze basal area, density, regeneration status and important value index. A total of 44 plant species were identified and classified into 41 genera and 28 families. The most relatively dominant species in the Hereje forest were Syzygium guineense (24.47%), Ficus sur (20%), Combretum molle (18.58%), Croton macrostachyus (18.11%), and Terminalia schimperiana (9.29%). According to this study, the density of tree species in the forest decreases as DBH classes increase. The predominance of small DBH classes is due to the excessive cutting of certain size classes by the local communities, indicating extensive changes in the woody plant diversity and age structure. The result of the study also revealed that diameter and height class distribution showed an inverted J-shape indicating the majority of the species had the highest number of individuals at lower DBH. Maytenus gracilipes has the highest number of seedlings in the study area. However, the overall regeneration status of the study forest was in fair condition. Hence, immediate conservation is required for these species with poor regeneration and least IVI should be given conservation priority.

Current Situation, Existing Problems and Prospect of the Study of Social Stratification in Contemporary China

Introduction. Under the backdrop of market economy and social transformation in China, the class structure of society has undergone profound changes, and the traditional class division is no longer relevant. However, there are still some unresolved problems in the study of social strata in China. The lack of diversity and comparability of research methods limits the depth and scope of research. Relatively few studies have been conducted on some emerging classes. Research on the relationship between social class and other social issues is insufficient, such as the relationship between class and education, class and income distribution, etc., are still limited. Methodology and sources. At present, a social classification model has been proposed to study social stratification in China, taking into account the characteristics of China's national conditions. A large number of empirical studies have revealed the evolution and characteristics of China's social class structure. The mechanism of social class formation and the factors influencing them have been studied, which has provided theoretical support for social development. Modern stratification models of Chinese authors have been used for the analysis in comparison with traditional models. Results and discussion. The study of social stratification in China inherited the theoretical framework and empirical research methods of Western social scientists, and there are also attempts to develop its own model of social stratification in combination with the actual situation in China. The structure of social stratification in China is characterized by the need to take into account the state redistribution system and the market economy model. However, there is a consensus that government officials are at the top of the social hierarchy, and the class of peasants and unemployed are at the bottom of society; mental workers (white collars) are higher than manual workers (blue collars). The results of studies on the impact of human capital on social stratification show that gender and educational level have universal significance in society; household registration system and housing type are special independent variables that affect social stratification in China more than in other countries, etc. However, there is a lack of relevant empirical studies that could provide specific quantitative data showing how independent variables affect social stratification in China. Many new social and professional groups have emerged in Chinese society. In recent years, some Chinese scholars have studied such specific social and professional groups, the most widespread and in demand in the labor market in China, taking into account the digitalization processes. Conclusion. It is necessary to further explore the methods and models of social stratification suitable for China's national conditions so as to cover the characteristics of social class diversity. It is necessary to strengthen the research on the relationship between social class and other social issues so as to provide more effective policy recommendations for social development. It is also necessary to strengthen the research on emerging classes, pay attention to their social status, mobility and responsibilities, and provide more effective support and guidance for their development.

LCANet: a model for analysis of students real-time sentiment by integrating attention mechanism and joint loss function

By recognizing students’ facial expressions in actual classroom situations, the students’ emotional states can be quickly uncovered, which can help teachers grasp the students’ learning rate, which allows teachers to adjust their teaching strategies and methods, thus improving the quality and effectiveness of classroom teaching. However, most previous facial expression recognition methods have problems such as missing key facial features and imbalanced class distributions in the dateset, resulting in low recognition accuracy. To address these challenges, this paper proposes LCANet, a model founded on a fused attention mechanism and a joint loss function, which allows the recognition of students’ emotions in real classroom scenarios. The model uses ConvNeXt V2 as the backbone network to optimize the global feature extraction capability of the model, and at the same time, it enables the model to pay closer attention to the key regions in facial expressions. We incorporate an improved Channel Spatial Attention (CSA) module as a way to extract more local feature information. Furthermore, to mitigate the class distribution imbalance problem in the facial expression dataset, we introduce a joint loss function. The experimental results show that our LCANet model has good recognition rates on both the public emotion datasets FERPlus, RAF-DB and AffectNet, with accuracies of 91.43%, 90.03% and 64.43%, respectively, with good robustness and generalizability. Additionally, we conducted experiments using the model in real classroom scenarios, detecting and accurately predicting students’ classroom emotions in real time, which provides an important reference for improving teaching in smart teaching scenarios.

Hybrid Oversampling and Undersampling Method (HOUM) via Safe-Level SMOTE and Support Vector Machine

The improvements in collecting and processing data using machine learning algorithms have increased the interest in data mining. This trend has led to the development of real-life decision support systems (DSSs) in diverse areas such as biomedical informatics, fraud detection, natural language processing, face recognition, autonomous vehicles, image processing, and each part of the real production environment. The imbalanced datasets in some of these studies, which result in low performance measures, have highlighted the need for additional efforts to address this issue. The proposed method (HOUM) is used to address the issue of imbalanced datasets for classification problems in this study. The aim of the model is to prevent the overfitting problem caused by oversampling and valuable data loss caused by undersampling in imbalanced data and obtain successful classification results. The HOUM is a hybrid approach that tackles imbalanced class distribution challenges, refines datasets, and improves model robustness. In the first step, majority-class data points that are distant from the decision boundary obtained via SVM are reduced. If the data are not balanced, SLS is employed to augment the minority-class data. This loop continues until the dataset becomes balanced. The main contribution of the proposed method is reproducing informative minority data using SLS and diminishing non-informative majority data using the SVM before applying classification techniques. Firstly, the efficiency of the proposed method, the HOUM, is verified by comparison with the SMOTE, SMOTEENN, and SMOTETomek techniques using eight datasets. Then, the results of the W-SIMO and RusAda algorithms, which were developed for imbalanced datasets, are compared with those of the HOUM. The strength of the HOUM is revealed through this comparison. The proposed HOUM algorithm utilizes a real dataset obtained from a project endorsed by The Scientific and Technical Research Council of Turkey. The collected data include quality control and processing parameters of yarn data. The aim of this project is to prevent yarn breakage errors during the weaving process on looms. This study introduces a decision support system (DSS) designed to prevent yarn breakage during fabric weaving. The high performance of the algorithm may encourage producers to manage yarn flow and enhance the HOUM’s efficiency as a DSS.