Classification Of Imbalanced Datasets Research Articles

Dimensionality reduction techniques in structural and earthquake engineering

Dimensionality reduction (DR) techniques are becoming increasingly important in engineering and scientific applications for feature extraction and data visualization purposes. This problem gets significantly more demanding when encountering class-imbalanced data sets with a disproportionate ratio of observations in each class. The main objective of this paper is to thoroughly evaluate the performance of various dimensionality reduction techniques for analyzing a cognate set of computational and experimental data in the field of structural and earthquake engineering with varying numbers of classes. Notably, a data set is generated using a computer simulation that characterizes risks posed by earthquakes to structures. Depending on the severity of damage in the simulations, data is classified into three, five, and eight classes. This research study also uses three existing multi-class imbalanced data sets that have been developed independently. For each problem, Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), and t-Distributed Stochastic Neighbor Embedding (t-SNE) are used, to cover a wide range of linear/non-linear and supervised/unsupervised DR techniques. To overcome the class-imbalance problem, the efficacy of the Synthetic Minority Oversampling (SMOTE) technique considering various classifiers, including K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Decision Trees (DT), Random Forest (RF), MLPClassifier (MLP), and AdaBoost are investigated. It is observed that combining SMOTE and LDA shows promising results for training classifiers on the reduced data. In light of the findings of this study, it is recommended to train DT classifiers after reducing the dimensionality of the input data. It is through this process that the scientific data is transformed into a two-dimensional space for visualizing decision surfaces, which is helpful for practitioners to understand the performance of classifiers.

A new hybrid approach based on genetic algorithm and support vector machine methods for hyperparameter optimization in synthetic minority over-sampling technique (SMOTE)

<abstract> <p>The crucial problem when applying classification algorithms is unequal classes. An imbalanced dataset problem means, particularly in a two-class dataset, that the group variable of one class is comparatively more dominant than the group variable of the other class. The issue stems from the fact that the majority class dominates the minority class. The synthetic minority over-sampling technique (SMOTE) has been developed to deal with the classification of imbalanced datasets. SMOTE algorithm increases the number of samples by interpolating between the clustered minority samples. The SMOTE algorithm has three critical parameters, "k", "perc.over", and "perc.under". "perc.over" and "perc.under" hyperparameters allow determining the minority and majority class ratios. The "k" parameter is the number of nearest neighbors used to create new minority class instances. Finding the best parameter value in the SMOTE algorithm is complicated. A hybridized version of genetic algorithm (GA) and support vector machine (SVM) approaches was suggested to address this issue for selecting SMOTE algorithm parameters. Three scenarios were created. Scenario 1 shows the evaluation of support vector machine SVM) results without using the SMOTE algorithm. Scenario 2 shows that the SVM was used after applying SMOTE algorithm without the GA algorithm. In the third scenario, the results were analyzed using the SVM algorithm after selecting the SMOTE algorithm's optimization method. This study used two imbalanced datasets, drug use and simulation data. After, the results were compared with model performance metrics. When the model performance metrics results are examined, the results of the third scenario reach the highest performance. As a result of this study, it has been shown that a genetic algorithm can optimize class ratios and k hyperparameters to improve the performance of the SMOTE algorithm.</p> </abstract>

A collaborative prediction approach to defend against amplified reflection and exploitation attacks

<abstract><p>An amplified reflection and exploitation-based distributed denial of service (DDoS) attack allows an attacker to launch a volumetric attack on the target server or network. These attacks exploit network protocols to generate amplified service responses through spoofed requests. Spoofing the source addresses allows attackers to redirect all of the service responses to the victim's device, overwhelming it and rendering it unresponsive to legitimate users. Mitigating amplified reflection and exploitation attacks requires robust defense mechanisms that are capable of promptly identifying and countering the attack traffic while maintaining the availability and integrity of the targeted systems. This paper presents a collaborative prediction approach based on machine learning to mitigate amplified reflection and exploitation attacks. The proposed approach introduces a novel feature selection technique called closeness index of features (CIF) calculation, which filters out less important features and ranks them to identify reduced feature sets. Further, by combining different machine learning classifiers, a voting-based collaborative prediction approach is employed to predict network traffic accurately. To evaluate the proposed technique's effectiveness, experiments were conducted on CICDDoS2019 datasets. The results showed impressive performance, achieving an average accuracy, precision, recall and F1 score of 99.99%, 99.65%, 99.28% and 99.46%, respectively. Furthermore, evaluations were conducted by using AUC-ROC curve analysis and the Matthews correlation coefficient (MCC) statistical rate to analyze the approach's effectiveness on class imbalance datasets. The findings demonstrated that the proposed approach outperforms recent approaches in terms of performance. Overall, the proposed approach presents a robust machine learning-based solution to defend against amplified reflection and exploitation attacks, showcasing significant improvements in prediction accuracy and effectiveness compared to existing approaches.</p></abstract>

Evolutionary Simultaneous Under and Oversampling of Instances for Dealing with Class-Imbalance Datasets in Multilabel Problems

Evolutionary Simultaneous Under and Oversampling of Instances for Dealing with Class-Imbalance Datasets in Multilabel Problems

Imbalanced COVID-19 dataset classification with bidirectional sampling based on sample correlation

Imbalanced COVID-19 dataset classification with bidirectional sampling based on sample correlation

Imbalanced COVID-19 dataset classification with bidirectional sampling based on sample correlation

Imbalanced COVID-19 dataset classification with bidirectional sampling based on sample correlation

EfficientNetV2 Model for Plant Disease Classification and Pest Recognition

Plant disease classification and prevention of spreading of the disease at earlier stages based on visual leaves symptoms and Pest recognition through deep learning-based image classification is in the forefront of research. To perform the investigation on Plant and pest classification, Transfer Learning (TL) approach is used on EfficientNet-V2. TL requires limited labelled data and shorter training time. However, the limitation of TL is the pre-trained model network’s topology is static and the knowledge acquired is detrimentally overwriting the old parameters. EfficientNet-V2 is a Convolutional Neural Network (CNN) model with significant high speed learning rates across variable sized datasets. The model employs a form of progressive learning mechanism which expands the network topology gradually over the course of training process improving the model’s learning capacity. This provides a better interpretability of the model’s understanding on the test domains. With these insights, our work investigates the effectiveness of EfficienetV2 model trained on a class imbalanced dataset for plant disease classification and pest recognition by means of combining TL and progressive learning approach. This Progressive Learning for TL (PL-TL) is used in our work consisting of 38 classes of PlantVillage dataset of crops and fruit species, 5 classes of cassava leaf diseases and another dataset with around 102 classes of crop pest images downloaded from popular dataset platforms, though it is not a benchmark dataset. To test the predictability rate of the model in classifying leaf diseases with similar visual symptoms, Mix-up data augmentation technique is used at the ratio of 1:4 on corn and tomato classes which has high probability of misinterpretation of disease classes. Also, the paper compares the TL approach performed on the above mentioned three types of data set using well established CNN based Inceptionv3, and Vision Transformer a non-CNN model. It clearly depicts that EfficientNetV2 has an outstanding performance of 99.5%, 97.5%, 80.1% on Cassava, PlantVillage and IP102 datasets respectively at a faster rate irrespective of the data size and class distribution as compared to Inception-V3 and ViT models.The performance metrics in terms of accuracy, precision, f1-score is also studied.

Improving performance of decision threshold moving-based strategies by integrating density-based clustering technique

<abstract> <p>Class imbalance learning (CIL), which aims to addressing the performance degradation problem of traditional supervised learning algorithms in the scenarios of skewed data distribution, has become one of research hotspots in fields of machine learning, data mining, and artificial intelligence. As a postprocessing CIL technique, the decision threshold moving (DTM) has been verified to be an effective strategy to address class imbalance problem. However, no matter adopting random or optimal threshold designation ways, the classification hyperplane could be only moved parallelly, but fails to vary its orientation, thus its performance is restricted, especially on some complex and density variable data. To further improve the performance of the existing DTM strategies, we propose an improved algorithm called CDTM by dividing majority training instances into multiple different density regions, and further conducting DTM procedure on each region independently. Specifically, we adopt the well-known DBSCAN clustering algorithm to split training set as it could adapt density variation well. In context of support vector machine (SVM) and extreme learning machine (ELM), we respectively verified the effectiveness and superiority of the proposed CDTM algorithm. The experimental results on 40 benchmark class imbalance datasets indicate that the proposed CDTM algorithm is superior to several other state-of-the-art DTM algorithms in term of G-mean performance metric.</p> </abstract>

Efficient Feature Evaluation Approach for a class-imbalanced dataset using Machine learning

Intrusion detection systems are a prominent field of research in order to identify attacks on computer networks. Data packets have many dimensions; therefore, examining them takes time. These dimensions include certain unimportant noises which are removed during pre-processing phase to increase the performance. Feature shuffle technique with random forest to calculate the roc-values of all the features used. A common machine learning algorithm for intrusion detection is used for implementation. Benchmark datasets CIC-DDoS2019 and Edge-IIoT were used to validate the proposed IDS. The experiment findings show that the proposed model is more accurate and has a higher Matthews Correlation Coefficient (MCC).

Two-Branch Generative Adversarial Network With Multiscale Connections for Hyperspectral Image Classification

Hyperspectral image (HSI) classification has always drawn great attention in the field of remote sensing. Various deep learning models are in the ascendant and gradually applied to HSI classification. Nevertheless, limited-labeled and class-imbalanced datasets largely make the classifier prone to overfitting. To address the above problem, this article proposes a two-branch generative adversarial network with multiscale connections (TBGAN), which includes two generators to produce the spectral and spatial samples, respectively. Thereinto, the spectral generator is imbued with the self-attention mechanism to maximumly capture the long-term dependencies across the spectral bands. And meanwhile, an elaborated discriminator with two branches is devised in TBGAN for extracting the joint spectral-spatial features. Besides, the multiscale connections are placed between the discriminator and two generators to alleviate the instability problems caused by the inherently backward propagation of gradients in GAN. Furthermore, a feature-matching term is added to the loss function to prevent the generators from overtraining upon the current discriminator, thereby further improving the stability of the network. Experiments upon three benchmark datasets demonstrate that TBGAN achieves an extremely competitive classification accuracy and exerts lower sensitivity to the training sample size compared with several state-of-the-art methods.

Deep-learning-based AI for evaluating estimated nonperfusion areas requiring further examination in ultra-widefield fundus images

We herein propose a PraNet-based deep-learning model for estimating the size of non-perfusion area (NPA) in pseudo-color fundus photos from an ultra-wide-field (UWF) image. We trained the model with focal loss and weighted binary cross-entropy loss to deal with the class-imbalanced dataset, and optimized hyperparameters in order to minimize validation loss. As expected, the resultant PraNet-based deep-learning model outperformed previously published methods. For verification, we used UWF fundus images with NPA and used Bland–Altman plots to compare estimated NPA with the ground truth in FA, which demonstrated that bias between the eNPA and ground truth was smaller than 10% of the confidence limits zone and that the number of outliers was less than 10% of observed paired images. The accuracy of the model was also tested on an external dataset from another institution, which confirmed the generalization of the model. For validation, we employed a contingency table for ROC analysis to judge the sensitivity and specificity of the estimated-NPA (eNPA). The results demonstrated that the sensitivity and specificity ranged from 83.3–87.0% and 79.3–85.7%, respectively. In conclusion, we developed an AI model capable of estimating NPA size from only an UWF image without angiography using PraNet-based deep learning. This is a potentially useful tool in monitoring eyes with ischemic retinal diseases.

Seizure onset zone identification using phase-amplitude coupling and multiple machine learning approaches for interictal electrocorticogram

Automatic seizure onset zone (SOZ) localization using interictal electrocorticogram (ECoG) improves the diagnosis and treatment of patients with medically refractory epilepsy. This study aimed to investigate the characteristics of phase-amplitude coupling (PAC) extracted from interictal ECoG and the feasibility of PAC serving as a promising biomarker for SOZ identification. We employed the mean vector length modulation index approach on the 20-s ECoG window to calculate PAC features between low-frequency rhythms (0.5–24 Hz) and high frequency oscillations (HFOs) (80–560 Hz). We used statistical measures to test the significant difference in PAC between the SOZ and non-seizure onset zone (NSOZ). To overcome the drawback of handcraft feature engineering, we established novel machine learning models to learn automatically the characteristics of the obtained PAC features and classify them to identify the SOZ. Besides, to handle imbalanced dataset classification, we introduced novel feature-wise/class-wise re-weighting strategies in conjunction with classifiers. In addition, we proposed a time-series nest cross-validation to provide more accurate and unbiased evaluations for this model. Seven patients with focal cortical dysplasia were included in this study. The experiment results not only showed that a significant coupling at band pairs of slow waves and HFOs exists in the SOZ when compared with the NSOZ, but also indicated the effectiveness of the PAC features and the proposed models in achieving better classification performance .

A deep residual neural network for semiconductor defect classification in imbalanced scanning electron microscope datasets

The detection of defects using inspection systems is common in a wide range of corporations such as semiconductor industries. The use of techniques based on deep learning (DL) and, in particular, convolutional neural networks (CNNs), emerges as a powerful tool to classify aesthetic defects and other unwanted anomalies in industrial automation applications marked by their natural complexity and high degree of variability. In this paper, an effective hybrid deep residual neural network approach is presented which merges traditional computer vision techniques to perform an appropriate defect segmentation and a deep residual neural network-grid search-based hyperparameter optimization for defect classification. The proposed model has been compared with other baseline algorithms and with other hybrid methods-based CNNs using different performance metrics such as F1-score, Cohen’s kappa coefficient, confusion matrix and computing time, on imbalanced datasets obtained from scanning electron microscope (SEM) images. The results obtained illustrate that the designed hybrid method provides the best defect classification of defects in semiconductor wafers in terms of F1-score (99.443%) while consuming the least computational time.

A novel hybrid resampling for semiconductor wafer defect bin classification

AbstractGenerally, defective dies on semiconductor wafer maps tend to form spatial clusters in distinguishable patterns which contain crucial information on specific problems of equipment or process, thus it is highly important to identify and classify diverse defect patterns accurately. However, in practice, there exists a serious class imbalance problem, that is, the number of the defective dies on semiconductor wafer maps is usually much smaller than that of the non‐defective dies. In various machine learning applications, a typical classification algorithm is, however, developed under the assumption that the number of instances for each class is nearly balanced. If the conventional classification algorithm is applied to a class imbalanced dataset, it may lead to incorrect classification results and degrade the reliability of the classification algorithm. In this research, we consider the semiconductor wafer defect bin data combined with wafer warpage information and propose a new hybrid resampling algorithm to improve performance of classifiers. From the experimental analysis, we show that the proposed algorithm provides better classification performance compared to other data preprocessing methods regardless of classification models.

Inverse free reduced universum twin support vector machine for imbalanced data classification

Imbalanced datasets are prominent in real-world problems. In such problems, the data samples in one class are significantly higher than in the other classes, even though the other classes might be more important. The standard classification algorithms may classify all the data into the majority class, and this is a significant drawback of most standard learning algorithms, so imbalanced datasets need to be handled carefully. One of the traditional algorithms, twin support vector machines (TSVM), performed well on balanced data classification but poorly on imbalanced datasets classification. In order to improve the TSVM algorithm’s classification ability for imbalanced datasets, recently, driven by the universum twin support vector machine (UTSVM), a reduced universum twin support vector machine for class imbalance learning (RUTSVM) was proposed. The dual problem and finding classifiers involve matrix inverse computation, which is one of RUTSVM’s key drawbacks. In this paper, we improve the RUTSVM and propose an improved reduced universum twin support vector machine for class imbalance learning (IRUTSVM). We offer alternative Lagrangian functions to tackle the primal problems of RUTSVM in the suggested IRUTSVM approach by inserting one of the terms in the objective function into the constraints. As a result, we obtain new dual formulation for each optimization problem so that we need not compute inverse matrices neither in the training process nor in finding the classifiers. Moreover, the smaller size of the rectangular kernel matrices is used to reduce the computational time. Extensive testing is carried out on a variety of synthetic and real-world imbalanced datasets, and the findings show that the IRUTSVM algorithm outperforms the TSVM, UTSVM, and RUTSVM algorithms in terms of generalization performance.

Synthesizing realistic high-resolution retina image by style-based generative adversarial network and its utilization

Realistic image synthesis based on deep learning is an invaluable technique for developing high-performance computer aided diagnosis systems while protecting patient privacy. However, training a generative adversarial network (GAN) for image synthesis remains challenging because of the large amounts of data required for training various kinds of image features. This study aims to synthesize retinal images indistinguishable from real images and evaluate the efficacy of the synthesized images having a specific disease for augmenting class imbalanced datasets. The synthesized images were validated via image Turing tests, qualitative analysis by retinal specialists, and quantitative analyses on amounts and signal-to-noise ratios of vessels. The efficacy of synthesized images was verified by deep learning-based classification performance. Turing test shows that accuracy, sensitivity, and specificity of 54.0 ± 12.3%, 71.1 ± 18.8%, and 36.9 ± 25.5%, respectively. Here, sensitivity represents correctness to find real images among real datasets. Vessel amounts and average SNR comparisons show 0.43% and 1.5% difference between real and synthesized images. The classification performance after augmenting synthesized images outperforms every ratio of imbalanced real datasets. Our study shows the realistic retina images were successfully generated with insignificant differences between the real and synthesized images and shows great potential for practical applications.

An End-to-End Cardiac Arrhythmia Recognition Method with an Effective DenseNet Model on Imbalanced Datasets Using ECG Signal.

Electrocardiography (ECG) is a well-known noninvasive technique in medical science that provides information about the heart's rhythm and current conditions. Automatic ECG arrhythmia diagnosis relieves doctors' workload and improves diagnosis effectiveness and efficiency. This study proposes an automatic end-to-end 2D CNN (two-dimensional convolution neural networks) deep learning method with an effective DenseNet model for addressing arrhythmias recognition. To begin, the proposed model is trained and evaluated on the 97720 and 141404 beat images extracted from the Massachusetts Institute of Technology-Beth Israel Hospital (MIT-BIH) arrhythmia and St. Petersburg Institute of Cardiological Technics (INCART) datasets (both are imbalanced class datasets) using a stratified 5-fold evaluation strategy. The data is classified into four groups: N (normal), V (ventricular ectopic), S (supraventricular ectopic), and F (fusion), based on the Association for the Advancement of Medical Instrumentation® (AAMI). The experimental results show that the proposed model outperforms state-of-the-art models for recognizing arrhythmias, with the accuracy of 99.80% and 99.63%, precision of 98.34% and 98.94%, and F1-score of 98.91% and 98.91% on the MIT-BIH arrhythmia and INCART datasets, respectively. Using a transfer learning mechanism, the proposed model is also evaluated with only five individuals of supraventricular MIT-BIH arrhythmia and five individuals of European ST-T datasets (both of which are also class imbalanced) and achieved satisfactory results. So, the proposed model is more generalized and could be a prosperous solution for arrhythmias recognition from class imbalance datasets in real-life applications.

Microclustering-Based Multi-Class Classification on Imbalanced Multi-Relational Datasets

In a relational database, classification algorithms are used to look for patterns across several interconnected relations. Most of the methods for multi-relational classification algorithms implicitly assume that the classes in the target relation are equally represented. Thus, they tend to produce poor predictive performance over the imbalanced dataset. In this paper, the authors propose an algorithm-level method MCMRC_IB for the classification of imbalanced multi-relational dataset. The proposed method extends MCMRC which is for balanced datasets. MCMRC_IB exploits the property of the imbalanced datasets that the minority class is represented by a smaller number of records usually 20-30% of the total records and is to be dealt accordingly by giving them weightages. The proposed method is able to handle multiple classes. Experimental results conðrm the efficiency of the proposed method in terms of predictive accuracy, F-measure, and G-mean.

Hybrid selection framework for class balancing approaches based on integrated CNN and decision making techniques for lung cancer diagnosis

Lung cancer is the fastest-growing and most dangerous type of cancer worldwide. It ranks first among cancer diseases in the number of deaths, and diagnosing it at late stages makes treatment more difficult. Artificial intelligence has played an essential role in the medical field in general, and early diagnosis of diseases and analyzing medical images in particular, as it can reduce human errors that may occur with the medical expert in medical image analysis. In this study, a hybrid framework is proposed between deep learning using the proposed convolutional neural network and multi-criteria decision-making techniques in order to reach an effective and accurate classification model for lung cancer diagnosis and select the best methodology to solve the problem of class imbalance datasets, which is a general problem in medical data that causes problems and errors in prediction. The IQ-OTHNCCD dataset that has a class imbalance was used. Three class balancing techniques were used separately and the data from each one enters the proposed convolutional neural network for feature extraction and classification. Then the Fuzzy-Weighted Zero-Inconsistency algorithm and VIKOR were used to make the ranking for the best classification approach and determine the best technique to balance the classes. This contributed to increasing the efficiency of the classification, where the best model got an accuracy of 99.27 %, sensitivity of 99.33 %, specificity of 99 %, precision of 98.67 % and F1-score of 99 %. This study can be applied to any data that suffers from the class imbalance problem to find the best technique that gives the highest classification accuracy.

Class-Difficulty Based Methods for Long-Tailed Visual Recognition

Long-tailed datasets are very frequently encountered in real-world use cases where few classes or categories (known as majority or head classes) have higher number of data samples compared to the other classes (known as minority or tail classes). Training deep neural networks on such datasets gives results biased towards the head classes. So far, researchers have come up with multiple weighted loss and data re-sampling techniques in efforts to reduce the bias. However, most of such techniques assume that the tail classes are always the most difficult classes to learn and therefore need more weightage or attention. Here, we argue that the assumption might not always hold true. Therefore, we propose a novel approach to dynamically measure the instantaneous difficulty of each class during the training phase of the model. Further, we use the difficulty measures of each class to design a novel weighted loss technique called ‘class-wise difficulty based weighted (CDB-W) loss’ and a novel data sampling technique called ‘class-wise difficulty based sampling (CDB-S)’. To verify the wide-scale usability of our CDB methods, we conducted extensive experiments on multiple tasks such as image classification, object detection, instance segmentation and video-action classification. Results verified that CDB-W loss and CDB-S could achieve state-of-the-art results on many class-imbalanced datasets such as ImageNet-LT, LVIS and EGTEA, that resemble real-world use cases.