Imbalanced Data Classification Research Articles

Domain adaptation with label-aligned sampling (DALAS) for cross-domain fault diagnosis of rotating machinery under class imbalance

Though existing cross-domain fault-diagnosis methods have shown promising results under domain shift conditions, existing approaches are only valid for class-balanced data. However, situations of class imbalance are inevitable in industrial fields, due to the difficulty in acquiring fault data from in-use machinery. Thus, if existing cross-domain fault diagnosis approaches are directly applied when domain shift and class imbalance coexist, performance degradation can occur. Thus, this research develops a domain adversarial learning network for class-imbalanced data to address situations where class imbalance and domain shift coexist; this situation is referred to as the problem of class imbalance domain adaptation (CIDA). In the proposed method, domain adversarial training is implemented for learning domain-invariant features by reducing the domain shift, and a label-aligned sampling strategy is utilized to deal with the class imbalance. In addition, for further performance enhancement of label-aligned sampling by increasing the accuracy of pseudo labels, metric learning is introduced to enhance the feature distinctiveness by expanding the distance of samples from different classes while decreasing the distance of samples from the same class. The efficiency of the proposed method is validated by applying it to various circumstances using two bearing datasets. The proposed method demonstrates superior performance compared to conventional algorithms in addressing the CIDA problem, according to quantitative and qualitative evaluations.

VGAN-BL: imbalanced data classification based on generative adversarial network and biased loss

VGAN-BL: imbalanced data classification based on generative adversarial network and biased loss

HSNF: Hybrid sampling with two-step noise filtering for imbalanced data classification

Imbalanced data classification has received much attention in machine learning, and many oversampling methods exist to solve this problem. However, these methods may suffer from insufficient noise filtering, overlap between synthetic and original samples, etc., resulting in degradation of classification performance. To this end, we propose a hybrid sampling with two-step noise filtering (HSNF) method in this paper, which consists of three modules. In the first module, HSNF denoises twice according to different noise discrimination mechanisms. Note that denoising mechanism is essentially based on the Euclidean distance between samples. Then in the second module, the minority class samples are divided into two categories, boundary samples and safe samples, respectively, and a portion of the boundary majority class samples are removed. In the third module, different oversampling methods are used to synthesize instances for boundary minority class samples and safe minority class samples. Experimental results on synthetic data and benchmark datasets demonstrate the effectiveness of HSNF in comparison with several popular methods. The code of HSNF will be released.

Modeling of class imbalance handling with optimal deep learning enabled big data classification model

Big data is the amount of data that surpasses the ability to process the data of a system concerning memory usage and computation time. It is commonly applied in several domains like healthcare, education, social networks, e-commerce, etc., as they have progressively obtained a massive quantity of input data. A major research problem is big data analytics, which can be carried out using expert systems and deep structured architectures. Besides, data wrangling and class imbalance data handling are challenging issues that need to be resolved in big data analytics. Class imbalance data degrade the performance of the classification model, which remains a challenging process due to the heterogeneous and complex structure of the comparatively huge datasets. Thus, the research focused on presenting a Class Imbalance Handling with Optimal Deep Learning Enabled Big Data Classification (CIHODL-BDC) framework. The core perception of the CIHODL-BDC framework helps to classify the big data in the Hadoop MapReduce framework. To accomplish this, the presented CIHODL-BDC model initially performs a data wrangling process is performed to alter the unrefined data into a useful layout. Next, the CIHODL-BDC model handles the class imbalance problem using a grey wolf optimizer (GWO) with Synthetic Minority Oversampling (SMOTE) technique. Besides, the Adam optimizer procedure with the Bidirectional Long Short Term Memory (BiLSTM) approach is performed to categorize the big data. The result analysis of the proposed CIHODL-BDC model is evaluated by two standard datasets. The simulation outcomes revealed the elevated performance of the CIHODL-BDC approach over existing methods.

Chinese relation extraction in military field based on multi-grained lattice transformer and imbalanced data classification

Relation extraction (RE) is a crucial step for knowledge graph construction, which aims to extract meaningful relations between entity pairs in plain texts. Very few works have been studied on Chinese relation extraction (CRE) in the military field. Moreover, recent deep neural network-based methods have achieved considerable performance but still suffer from three inherent limitations, including overlapping of entities, imbalanced data and the ambiguity. Therefore, this work investigates a novel Multi-Grained Lattice Transformer (MGLT), which leverages external information of lexicon and word sense tailored for CRE. In MGLT, self-matched lexicon words and related word senses are fused through a cross-transformer mechanism to alleviate the ambiguity in texts. The finally enriched sequence representation in MGLT captures the relatedness between the head entity and the tail one, which is helpful to alleviate the overlapping of entities. Experimental results on two benchmark datasets and a self-developed dataset constructed from online military news show that the proposed MGLT achieves state-of-the-art (SOTA) performance. Compared with other typical baselines, MGLT achieves better area under curve (AUC) and [Formula: see text]-score by up to 10.46% and 6.90%, respectively. We further demonstrate the effectiveness of using ensemble learning to fully exploit complementary information from multiple MGLT-based base learners to improve the overall performance for imbalanced data classification on the military dataset. Such results indicate that the proposed ensemble learning model is effective and robust to be applied in practical applications.

Learning from class-imbalanced data using misclassification-focusing generative adversarial networks

This paper presents a novel end-to-end oversampling-classification approach, which we refer to as imbalanced data-classifying generative adversarial network (ImbGAN), for imbalanced data classification. ImbGAN has a classifier-embedded structure within a GAN and consists of five components: (1) generator, (2) discriminator, (3) classifier, (4) storage for misclassified minority class data, and (5) storage for artificial minority class data. By iterative interaction with the embedded classifier, the first two components generate artificial minority class instances that are similar to minority class instances misclassified by the classifier. Therefore, these three networks are iteratively and simultaneously trained. The misclassified and artificial minority class instances are stored in the fourth and fifth components, respectively. These two components are also updated as iterations proceed. Our method obtains the final classification model from a single learning process, while most artificial data generation methods for imbalanced data classification go through an additional process for training classifiers after artificial data generation. Numerical experiments based on tabular, image, and text datasets confirm that the proposed method outperforms well-known synthetic sampling methods.

Irrelevant attribute resistance approach to binary classification for imbalanced data

Imbalanced data distribution is a common feature in real-world datasets. For imbalanced data, the imbalanced characteristics of the classes have two negative effects on classification results, one of which is that the minority class is highly likely masked by the majority class so as to weaken the ability of the classifiers to identify the minority class. Another effect is that irrelevant attributes hidden in imbalanced data can create much noise to interfere the classifiers, thereby leading to that the classifiers could mistakenly treat noise as the minority classes. In this scenario, the performance of the classifiers is rapidly declined and the classifiers obtain incorrect classification results. To address this issue, this paper proposed a conformal transformation twin-hypersphere with fuzzy. The critical thought is that using conformal transformation to explore the regions containing minority classes, by so doing, minority classes can be more likely to be noticed by the classifier. Using the proposed fuzzy function assesses the contributions of points to the hypersphere training, through evaluating the contributions, noise can be determined, thereby increasing the ability of the classifier to noise resistance. Results on the synthetic and real datasets show that the proposed method outperforms the competitors in classification accuracy and noise resistance. Results also imply that the proposed method does not exhibit exponential calculation time, meaning that the method is suitable for the classification of large-scale imbalanced datasets. We demonstrate that conformal transformation can assist those non-linear kernels to find those hard-to-observe regions containing minority classes, thereby strengthening the adaptability of the classifiers to imbalanced data following complex distributions. Moreover, the non-linear kernels using conformal transformation can adapt to the situation where different sub-regions in sample space require different nonlinearities.

Adaptive SV-Borderline SMOTE-SVM algorithm for imbalanced data classification

In recent years, imbalanced data classification has emerged as a challenging task. To address this issue, we propose an adaptive SV-Borderline SMOTE-SVM (Synthetic Minority Oversampling Technique-Support Vector Machine) algorithm, specifically designed to overcome the challenges associated with imbalanced data classification. The algorithm begins by mapping the dataset into the kernel space using SVM to identify the class boundary samples, known as support vectors (SVs). Subsequently, the neighbors of positive sample’s support vector (SV+) are calculated based on the kernel distance. Based on the class distribution of these neighbors, the SV+ samples are labeled as either “concave” or “convex”. Based on these labels, new samples are adaptively generated using two distinct calculation approaches for different labeled SV+ samples. To construct the SVM decision function without requiring the explicit expression of new samples in the kernel space, a Gram matrix is designed. Notably, all the processes ensure the credibility and reliability of the new samples. Additionally, the adaptive interpolation approach helps to ensure the security and diversity of new samples. Extensive experiments were conducted on a set of 50 KEEL datasets to evaluate the performance of our proposed method for imbalanced data classification. In experiments, our method achieved the highest G-mean score in 33 datasets and the highest F-values in 32 datasets. These results highlight the effectiveness and superiority of our proposed method compared to other approaches in addressing the challenges of imbalanced data classification.

Enhancing anomaly detection accuracy and interpretability in low-quality and class imbalanced data: A comprehensive approach

With the rapid advancements in smart meters, energy internet, and high-performance computing technologies, deep learning methods are increasingly used for detecting anomalies in building electricity consumption. However, these models grapple with challenges like outliers, missing data, and imbalanced categories. Moreover, these models are often considered as black boxes due to their lack of interpretability. This limitation impedes the practical deployment of deep learning methods. Focusing on these issues, this study presents an anomaly detection method aimed at improving anomaly detection accuracy and interpretability in low-quality and class imbalanced data. Initially, our method established data preprocessing models to handle with outliers and missing data. Subsequently, an unbalance elimination model, named GRU-MACGANS, was designed to augment the minority category data. The SA-CNN-BiGRU-FCN model, incorporating a knowledge transfer strategy, was implemented to identify anomalous behavior. Finally, the results of SA-CNN-BiGRU-FCN model were interpreted based on the contrastive explanations method. Our case study of a low-voltage distribution system demonstrates the effectiveness of our approach. Data preprocessing models and GRU-MACGANS enhance data quality and alleviate category imbalance issues in anomaly detection. The interpretable analysis provides insights into the ability of detection models to identify representative features of different electricity consumption behaviors for effective anomaly detection.

A Machine Learning Approach for Predicting the Occurrence of Asparaginase-Associated Pancreatitis in Pediatric Patients with Acute Lymphoblastic Leukemia

Introduction Children being treated for acute lymphoblastic leukemia (ALL) are frequently affected by asparaginase-associated pancreatitis. Additionally, pancreatitis is among the most troublesome and frequent side effects of asparaginase therapy and is a significant contributor to early drug discontinuation and poor outcomes. There are inadequate odds ratios for known risk factors, such as asparaginase dosage, advanced age, and single nucleotide polymorphisms, to predict pancreatitis occurrence. The goal of this study was to use machine learning to develop a predictive model for asparaginase-induced pancreatitis in pediatric ALL patients. Methods Data were collected from 711 patients who had childhood ALL and received asparaginase. Pancreatitis was defined as serum amylase and/or lipase levels greater than three times the upper limit of normal or acute pancreatitis on abdominal images. One month from the time of asparaginase administration for each patient was defined as one “timestep”, and when asparaginase was administered thereafter, it was defined as a new individual timestep. Each timestep was defined as one training case, and a case in which pancreatitis occurred at that timestep was defined as an event. Finally, 3193 training cases were defined in a total of 711 patients. The physical measurement results, prescription codes, blood test results, and blood transfusion history data were collected from electronic health records (HER) during the entire treatment period of the patients. Among these are age, body mass index, body surface area, gender, type of asparaginase (Native, Erwinia, or Pegylated), previous history of pancreatitis, cumulative number of asparaginase administrations, and asparaginase change history before the current time point. The results of 47 blood tests on the start date of asparaginase in each timestep were also used as predictive variables (Figure 1). Using logistic regression, Random forest, and XCBoost as machine learning methods, we assessed a model predicting asparaginase-associated pancreatitis through 5-fold cross-validation. Performance indicators such as area under the receiver operating characteristic curve (AUC) score, Precision Recall (PR) score, F0.5 score, and F2 score were employed to evaluate the binary classification of imbalanced data. The selection of the model was determined based on these two criteria. Results When considering the F(0.5+F2.0)/2 score as the basis for model selection, the logistic regression model demonstrated an AUC of 81% (PR 32.86%, F(0.5+F2.0)/2 score 23.23%). On the other hand, the XGboost model exhibited an AUC of 79% (PR 33.7%, F(0.5+F2.0)/2 score 32.07%), while the Random Forest model achieved an AUC of 84% (PR 33.34%, F(0.5+F2.0)/2 score 39.48%). Among these models, the Random Forest model demonstrated the highest predictive power. When the model was chosen using the PR score, the logistic regression model achieved an AUC of 80% (PR 34.97%, F(0.5+F2.0)/2 score 22.08%), whereas the XGboost model achieved an AUC of 79% (PR 31.58%, F(0.5+F2.0)/2 score 31.6%). Also, it was seen that the Random Forest model had the best performance across all metrics, with an AUC of 85%, a precision-recall (PR) score of 32.26%, and a F(0.5+F2.0)/2 score of 36.4% (Figure 2, left). According to Shapley values, it is evident that some parameters, namely greater lipase levels, higher cumulative asparaginase dosages, higher amylase levels, higher glucose levels, and older age, have significantly contributed to the occurrence of asparaginase-associated pancreatitis (Figure 2, right). Conclusions A machine learning model was employed to successfully forecast the occurrence of acute pancreatitis following the administration of asparaginase in pediatric patients with AAL. This study specifically focused on making predictions regarding pancreatitis within a month based on the test results obtained at the commencement of asparaginase treatment. This approach offers the potential for promptly predicting the development of pancreatitis. In further stages, following external validation and prospective observational clinical trials, the prediction model has the potential to be included in the EHR and serve as a Clinical Decision Support System (CDSS).

A Hybrid Resampling Approach for Multiclass Skewed Datasets and Experimental Analysis with Diverse Classifier Models

In real-life scenarios, imbalanced datasets pose a prevalent challenge for classification tasks, where certain classes are heavily underrepresented compared to others. To combat this issue, this article introduces DOSAKU, a novel hybrid resampling technique that combines the strengths of DOSMOTE and AKCUS algorithms. By integrating both oversampling and undersampling methods, DOSAKU significantly reduces the imbalance ratio of datasets, enhancing the performance of classifiers. The proposed approach is evaluated on multiple models employing different classifiers, and the results demonstrate its superiority over existing resampling measures, making it an effective solution for handling class imbalance challenges. DOSAKU's promising performance is a substantial contribution to the field of imbalanced data classification, as it offers a robust and innovative solution for improving predictive model accuracy and fairness in real-world applications where imbalanced datasets are common.

The Performance Comparison between C4.5 Tree and One-Dimensional Convolutional Neural Networks (CNN1D) with Tuning Hyperparameters for the Classification of Imbalanced Medical Data

The Performance Comparison between C4.5 Tree and One-Dimensional Convolutional Neural Networks (CNN1D) with Tuning Hyperparameters for the Classification of Imbalanced Medical Data

Adaptive fuzzy multi-neighborhood feature selection with hybrid sampling and its application for class-imbalanced data

For imbalanced data, classification efficiency degrades significantly due to the missing information for the positive class, and existing sampling schemes do not consider the distributions of samples. Additionally, the global parameters of fuzzy neighborhoods are set manually. These defects affect the effectiveness of classifier. To address these problems, we offer an adaptive fuzzy multi-neighborhood feature selection methodology with intercluster distance-based hybrid sampling for class-imbalanced data. First, the number of clusters can be defined in terms of the number of samples in the negative or positive class. The initial centers of the clusters are determined according to the number of clusters, and the dissimilarity and similarity measures are calculated by using the intercluster distances between samples. Then, the cluster center, fuzzy membership matrix, and intercluster distance are studied, and then the optimization objective function is designed. The hybrid sampling scheme can be used to combine the generated positive class samples and negative class samples and obtain a class-balanced system. Second, according to the sample distribution, the standard deviation and a set of adaptive fuzzy multi-neighborhood radii are designed. A fuzzy multi-neighborhood similarity relation is defined by introducing a Gaussian kernel model to obtain a fuzzy multi-neighborhood granule, and an improved fuzzy multi-neighborhood rough set model is provided. Uncertain measures of fuzzy neighborhood systems are evaluated by the positive region and dependency. Third, by integrating fuzzy dependence with fuzzy complementary condition entropy, fuzzy multi-neighborhood complementary mutual information is provided on two viewpoints of algebra and information. Finally, a heuristic feature subset selection methodology for imbalanced classification with hybrid sampling using fuzzy c-means clustering is studied to obtain this excellent set of features. Experiments on 26 imbalanced datasets show the effectiveness of our designed algorithm.

Class-imbalanced time series anomaly detection method based on cost-sensitive hybrid network

The time series is a structured data form. In practical engineering, most of the time series is in the form of class-imbalance, in which, abnormal data is rare or underrepresented. In this study, we present a cost sensitive hybrid network (CSHN) model to detect data anomaly in class-imbalanced time series. The proposed model adopts a two-flow structure that consist of a Convolutional Neural Network (CNN) and a Gated Recurrent Unit (GRU) network and combine with a cost-sensitive loss function. This model integrates the advantages of CNN and GRU. The former is provided with strong learning ability for local state features and the latter can extract the sequential features and force the decision logic of the network to pay more attention to the long-term dependence of the time series. The cost-sensitive loss function is used to solve the problem that the detection accuracy of the minority class will be inaccuracy caused by skewed data distribution. The simulation experiments are conducted on the speed and temperature datasets that come from real-world measurement and control engineering, as well as the UCR datasets. Compared to using the cross-entropy loss function, the using of the cost-sensitive loss function can improve G-means and F-measure by 3 %∼5% and 5 %∼9%, respectively, on the speed and temperature datasets. Experimental results demonstrate that our method is efficient for the class-imbalanced data sets, and the numerical evaluation indexes are superior to that of the comparison methods.

How Far Have We Progressed in the Sampling Methods for Imbalanced Data Classification? An Empirical Study

Imbalanced data are ubiquitous in many real-world applications, and they have drawn a significant amount of attention in the field of data mining. A variety of methods have been proposed for imbalanced data classification, and data sampling methods are more prevalent due to their independence from classification algorithms. However, due to the increasing number of sampling methods, there is no consensus about which sampling method performs best, and contradictory conclusions have been obtained. Therefore, in the present study, we conducted an extensive comparison of 16 different sampling methods with four popular classification algorithms, using 75 imbalanced binary datasets from several different application domains. In addition, four widely-used measures were employed to evaluate the corresponding classification performance. The experimental results showed that none of the employed sampling methods performed the best and stably across all the used classification algorithms and evaluation measures. Furthermore, we also found that the performance of the different sampling methods was usually affected by the classification algorithms employed. Therefore, it is important for practitioners and researchers to simultaneously select appropriate sampling methods and classification algorithms, for handling the imbalanced data problems at hand.

A sparrow search algorithm-optimized convolutional neural network for imbalanced data classification using synthetic minority over-sampling technique

Due to the large span of flight delay, numerous influencing factors and large number of flights, the collected flight delay data is characterized by large-scale and imbalance, which is considered as one of the largest challenges in balanced flight delay data classification by using conventional classification algorithms. To solve this problem, an adaptive multi-class classification approach of imbalanced flight delay data using synthetic minority over-sampling technique (SMOTE)-based convolutional neural network with sparrow search algorithm, namely SSA-LTCNN is proposed in this paper. Firstly, the SMOTE is used to reduce the imbalance between the classes of the flight delay data. Then, the qualitative analysis method is employed to determine the key parameters of the CNN, which are optimized by using SSA to establish an optimized CNN model. Finally, the optimized CNN model with more effective feature representation ability is employed to extract discriminative features from the balanced flight delay data automatically. The obtained features in the last fully connected layer of the optimized CNN model are used for flight delay classification. The MIT-BIH Arrhythmia Data is selected to verify the effectiveness of the proposed SSA-LTCNN method and the confusion matrix, classification accuracy, precision, recall and F1 score are also used to evaluate the performance of the proposed SSA-LTCNN method, which can get better performance than RF, DT, LR, CNN and QRSCNN. The actual flight delay data is processed to verify the effectiveness of the proposed classification approach, which can achieve higher accuracy than KNN, NB, UCNN, CNN, SSA-LCNN and SSA-TCNN in different time periods. The experimental results show that the SSA-LTCNN can fast and effectively realize the classification of MIT-BIH Arrhythmia Data and flight delay data.

Crack segmentation of imbalanced data: The role of loss functions

Loss functions, which govern a deep learning-based optimization process, have been widely investigated to handle the class imbalanced data issue in crack segmentation. However, their performance varies according to models and datasets, making it challenging to choose the most appropriate ones. To address this issue, the paper conducts a large-scale performance comparison of twelve commonly used loss functions on four benchmark datasets. Various aspects are considered using a statistical test-based ranking scheme, which integrates accuracy, sensitivity to threshold change, and varying degrees of imbalance severity for a comprehensive comparison. The results show that most loss functions achieve relatively similar accuracy on mildly imbalanced datasets, while weighted binary cross-entropy loss, Focal loss, Dice-based loss, and compound loss functions significantly outperform others as imbalance severity increases. In general, Focal Tversky loss function exhibits excellent performance in handling the imbalanced data issue.

Data-level Hybrid Strategy Selection for Disk Fault Prediction Model Based on Multivariate Gan

Data class imbalance is a common problem in classification problems, where minority class samples are often more important and more costly to misclassify in a classification task. Therefore, it is very important to solve the data class imbalance classification problem. The SMART dataset exhibits an evident class imbalance, comprising a substantial quantity of healthy samples and a comparatively limited number of defective samples. This dataset serves as a reliable indicator of the disc's health status. In this paper, we obtain the best balanced disk SMART dataset for a specific classification model by mixing and integrating the data synthesised by multivariate generative adversarial networks (GAN) to balance the disk SMART dataset at the data level; and combine it with genetic algorithms to obtain higher disk fault classification prediction accuracy on a specific classification model.

A Survey of Methods for Handling Disk Data Imbalance

Class imbalance exists in many classification problems, and since the data is designed for accuracy, imbalance in data classes can lead to classification challenges with a few classes having higher misclassification costs. The Backblaze dataset, a widely used dataset related to hard discs, has a small amount of failure data and a large amount of health data, which exhibits a serious class imbalance. This paper provides a comprehensive overview of research in the field of imbalanced data classification. The discussion is organized into three main aspects: data-level methods, algorithmic-level methods, and hybrid methods. For each type of method, we summarize and analyze the existing problems, algorithmic ideas, strengths, and weaknesses. Additionally, the challenges of unbalanced data classification are discussed, along with strategies to address them. It is convenient for researchers to choose the appropriate method according to their needs.

Imbalanced data classification using improved synthetic minority over-sampling technique

In data mining, deep learning and machine learning models face class imbalance problems, which result in a lower detection rate for minority class samples. An improved Synthetic Minority Over-sampling Technique (SMOTE) is introduced for effective imbalanced data classification. After collecting the raw data from PIMA, Yeast, E.coli, and Breast cancer Wisconsin databases, the pre-processing is performed using min-max normalization, cleaning, integration, and data transformation techniques to achieve data with better uniqueness, consistency, completeness and validity. An improved SMOTE algorithm is applied to the pre-processed data for proper data distribution, and then the properly distributed data is fed to the machine learning classifiers: Support Vector Machine (SVM), Random Forest, and Decision Tree for data classification. Experimental examination confirmed that the improved SMOTE algorithm with random forest attained significant classification results with Area under Curve (AUC) of 94.30%, 91%, 96.40%, and 99.40% on the PIMA, Yeast, E.coli, and Breast cancer Wisconsin databases.