Unbalanced Class Distribution Research Articles

Mild cognitive impairment prediction and cognitive score regression in the elderly using EEG topological data analysis and machine learning with awareness assessed in affective reminiscent paradigm.

The main objective of this study is to evaluate working memory and determine EEG biomarkers that can assist in the field of health neuroscience. Our ultimate goal is to utilize this approach to predict the early signs of mild cognitive impairment (MCI) in healthy elderly individuals, which could potentially lead to dementia. The advancements in health neuroscience research have revealed that affective reminiscence stimulation is an effective method for developing EEG-based neuro-biomarkers that can detect the signs of MCI. We use topological data analysis (TDA) on multivariate EEG data to extract features that can be used for unsupervised clustering, subsequent machine learning-based classification, and cognitive score regression. We perform EEG experiments to evaluate conscious awareness in affective reminiscent photography settings. We use EEG and interior photography to distinguish between healthy cognitive aging and MCI. Our clustering UMAP and random forest application accurately predict MCI stage and MoCA scores. Our team has successfully implemented TDA feature extraction, MCI classification, and an initial regression of MoCA scores. However, our study has certain limitations due to a small sample size of only 23 participants and an unbalanced class distribution. To enhance the accuracy and validity of our results, future research should focus on expanding the sample size, ensuring gender balance, and extending the study to a cross-cultural context.

Network Traffic Classification Based on SD Sampling and Hierarchical Ensemble Learning

With the increase in cyber threats in recent years, there have been more forms of demand for network security protection measures. Network traffic classification technology is used to adapt to the dynamic threat environment. However, network traffic has a natural unbalanced class distribution problem, and the single model leads to the low accuracy and high false-positive rate of the traditional detection model. Given the above two problems, this paper proposes a new dataset balancing method named SD sampling based on the SMOTE algorithm. Different from the SMOTE algorithm, this method divides the sample into two types that are easy and difficult to classify and only balances the difficult-to-classify sample, which not only overcomes the SMOTE’s overgeneralization but also combines the idea of oversampling and undersampling. In addition, a two-layer structure combined with XGBoost and the random forest is proposed for multiclassification of anomalous traffic, since using a hierarchical structure can better classify minority abnormal traffic. This paper conducts experiments on the CICIDS2017 dataset. The results show that the classification accuracy of the proposed model is more than 99.70% and that the false-positive rate is less than 0.34%, indicating that the proposed model is better than traditional models.

Parallel Convolutional Neural Networks and Transfer Learning for Classifying Landforms in Satellite Images

The use of remote sensing has great potential for detecting many natural differences, such as disasters, climate changes, and urban changes. Due to technological advances in imaging, remote sensing has become an increasingly popular topic. One of the significant benefits of technological advancement has been the ease with which remote sensing data is now accessible. Physical and spatial information is detected by remote sensing, which can be described as the process of identifying distinctive characteristics of an environment. Resolution is one of the most important factors influencing the success of the detection processes. As a result of the resolution being below the necessary level, features of the objects to be differentiated become incomprehensible and therefore constitute a significant barrier to differentiation. The use of deep learning methods for classifying remote sensing data has become prevalent and successful in recent years. This study classified Satellite images using deep learning and machine learning methods. Based on the transfer learning strategy, a parallel convolutional neural network (CNN) was designed in the study. To improve the feature mapping of an image, convolutional branches use pre-trained knowledge of the transmitted network. Using the offline augmentation method, the raw data set was balanced to overcome its unbalanced class distribution and increased network performance. A total of 35 classes of landforms have been studied in the experiments. The accuracy value of the developed model in the classification study of landforms was 97.84%. According to experimental results, the proposed method provides high classification accuracy in detecting landforms and outperforms existing studies.

Label Propagation and Contrastive Regularization for Semisupervised Semantic Segmentation of Remote Sensing Images

Remarkable progress based on deep neural networks has been achieved on the semantic segmentation in remote sensing images. However, pixel-level labeling is expensive for remote sensing images. Semi-supervised semantic segmentation becomes an alternative approach to reduce the cost of annotation, and it is crucial to utilize efficiently a large number of unlabeled data. Nevertheless inevitably, there is the unbalanced class distribution between labeled and unlabeled data of remote sensing scene. Existing semi-supervised methods train unlabeled images in isolation from labeled images and only learn reliable pixel pseudo-labels, leading to underutilization of unlabeled images. This article proposes a novel semi-supervised semantic segmentation approach based on label propagation and contrastive regularization for remote sensing images. Specifically, the unlabeled images are augmented by randomly copy-pasting the class regions from labeled images. A prototype feature constraint module is used to enforce the constraint on the pixel features of unlabeled images relying on the prototype features from labeled images, achieving feature alignment on the entire dataset. Furthermore, we present the region contrastive learning module that guides the model to learn feature consistency under different perturbations and compact feature representations over class regions on unlabeled images. Extensive experimental results on multiple remote sensing datasets demonstrate that our proposed approach achieves superior performance compared with state-of-the-art semi-supervised semantic segmentation methods.

Comparative study on the performance of deep learning implementation in the edge computing: Case study on the plant leaf disease identification

This research examines and explores four different pre-trained CNN deep learning models (AlexNet, VGG16, ResNet50, and DenseNet121) to be adopted as edge solution. The model is developed and evaluated using the PlantVillage dataset. Image transformation techniques and down sampling were carried out to mitigate the unbalanced class distribution problem. From the preliminary works, DenseNet121 was selected as the implementation model since it had the best accuracy (96.4%) in comparison with other models. The model is then tested on different endpoint devices (CPU, GPU and VPU) in different programming environments (standard PyTorch and OpenVINO) to test the consistency of models that work across different hardware and software configurations. Results show that the adoption of model on the edge (VPU) is able to maintain relatively high recall values, precision and F1 scores. The successful application of the model on VPU device shows the potential of the model to be implemented for the detection of plant diseases as edge solution/embedded application. • Plant disease identification deep learning model has been developed. • Model performance in different hardware and environment settings has been compared. • The model in edge setting with VPU has a better accuracy and inference time.

Performance Enhancement of the Unbalanced Text Classification Problem Through a Modified Chi Square-Based Feature Selection Technique

This paper proposes a modified chi square-based feature selection algorithm in conjunction with a random vector functional link network-based text classifier for improving the classification performance of multi-labeled text documents with unbalanced class distributions. In the proposed feature selection method, maximum features are selected from classes that have a great deal of training and testing documents as an improvement towards original chi-square method. On two benchmark datasets that are multi-labeled, multi-class, and unbalanced, a comparison of the model with three conventional selection techniques such as chi-square, term frequency-inverse document frequency, and mutual information is accumulated for assessing its effectiveness. Additionally, the proposed model is compared with four different classifiers. In the study, it was found that the proposed model performs better in terms of precision, recall, f-measure, and hamming losses and is able to select the majority of true positive documents despite an unbalanced class distribution for both the datasets.

Generative adversarial networks for extrapolation of corrosion in automobile images

Deep learning has shown success in several applications involving pattern recognition, expert systems, and scientific discovery. However, existing methods struggle with industrial applications, which are often challenged by non-ideal datasets. In many cases, the datasets are small, poorly labeled, noisy, or have unbalanced class distribution, or any combination of such problems. In this contribution, we propose a generative adversarial network (GAN) strategy that is able to circumvent limitations imposed by very small datasets. As case study, we use the extrapolation of corrosion in automobiles and feed our deep learning framework with only few dozen images, as opposed to the thousands to million images commonly found in many computer vision problems. In order to handle such a reduced dataset, we use one GAN for the rust level and one for the rust texture. The rust level GAN is conditional on random samples from the dataset and uses an additive random noise in the latent space to add variability to the generated rust level maps. The rust texture GAN adds the shades of brown to the outputs of the rust level GAN. To improve the robustness of our approach, the rust level GAN is a conditional GAN with random noise added in the latent space of the generator; while the rust texture GAN uses supervised loss functions coming from the observed data set as well as an unsupervised loss function coming from the rust level GAN. In addition, given the very reduced size of our dataset, it is unfeasible to break down the data into training, validation, and test sets. We overcome this limitation by using the discrepancy between the generated and target distributions of the rust level and texture intensities as a way to monitor convergence of training. The resulting models are able to ingest an image with a car having no corrosion and generates an image of this car with parts exhibiting varying degrees of corrosion (from mild, to moderate, to severe). The source codes and links to the data can be found in the following GitHub repository https://github.com/PML-UCF/rusty_fender_gan.

Research on prediction of multi-class theft crimes by an optimized decomposition and fusion method based on XGBoost

The number of theft cases is much higher than that of other criminal cases, which frequently occurs in daily life and is seriously destructive to social order. Studying the law of theft cases has a positive impact on social governance and optimizing police deployment. Therefore, based on the data of theft cases in H city, this study proposes an optimized decomposition and fusion method based on XGBoost, and establishes two multi-classification prediction models, such as OVR-XGBoost and OVO-XGBoost. As the theft data is a datasets with unbalanced class distribution, this paper uses SMOTENN algorithm to process it into a datasets with balanced distribution, which effectively improves the effect of the model. Experiments show that the prediction accuracy of OVR-XGBoost and OVO-XGBoost models is higher than that of baseline XGBoost models. For categories with few samples, the classification effect of OVO-XGBoost is better than that of baseline XGBoost and OVO-XGBoost models. Compared with baseline XGBoost model, the average overall classification accuracy of OVO-XGBoost model is improved by more than 7%, and the MacroR accuracy is also improved by more than 15%. The model proposed in this study has a good effect on the classification and prediction of theft types, and is of great significance for the prevention of theft cases.

Comparative analysis of resampling techniques on Machine Learning algorithm

Generally, classification algorithms in the field of data science assume that the classes of training data are equally distributed. However, datasets on real problems often have an unbalanced class distribution. Unbalanced dataset classes make up the majority class and the minority class. In general, minority classes are more attractive and more important to identify. In this case, the correct classification for the minority class sample is more valuable than the majority class. The unbalanced class distribution causes the classification algorithm to have difficulty in classifying minority class samples correctly. If the performance of the algorithm model is good for the majority class sample but bad for the minority class then this imbalance problem is a crucial thing to be addressed. Many solutions are offered for this problem, namely by oversampling techniques in the minority class and/or undersampling techniques in the majority class. In this study, the authors tried various sampling techniques and tested them on various machine learning classification algorithms to find out the combination of resampling techniques and algorithms that have high recall in classifying minority class samples and still considering the majority class classification.

The Utility of Receiver Operating Characteristic Curve in Educational Assessment: Performance Prediction

When examinees are classified into groups based on scores from educational assessment, two indices are widely used to gauge the psychometric quality of the classifications: accuracy and consistency. The two indices take correct classifications into consideration while overlooking incorrect ones, where unbalanced class distribution threatens the validity of results from the accuracy and consistency indices. The single values produced from the two indices also fail to address the inconsistent accuracy of the classifier across different cut score locations. The current study proposed the concept of classification quality, which utilizes the receiver operating characteristics (ROC) graph to comprehensively evaluate the performance of classifiers. The ROC graph illustrates the tradeoff between benefits (true positive rate) and costs (false positive rate) in classification. In this article, a simulation study was conducted to demonstrate how to generate and interpret ROC graphs in educational assessment and the benefits of using ROC graphs to interpret classification quality. The results show that ROC graphs provide an efficient approach to (a) visualize the fluctuating performance of scoring classifiers, (b) address the unbalanced class distribution issue inherent in the accuracy and consistency indices, and (c) produce more accurate estimation of the classification results.

Effective predictive modelling for coronary artery diseases using support vector machine

Coronary artery disease (CAD) is a category of cardiovascular disease that causes the highest mortality rate in the world. CAD occurs due to plaque build-up on the walls of the arteries that supply blood to the heart and other organs of the body. To control the mortality rate, a practical model that is capable of predicting CAD is needed. Machine learning approaches have been used in solving various problems in various domains, including biomedicine. However, real-world data often has an unbalanced class distribution that can interfere with classifier performance. In addition, data has many features to process. This study focuses on effective modeling capable of predicting CAD using feature selection to handle high dimensional data and feature resampling to handle unbalanced data. Feature selection is very effective by eliminating irrelevant features from the training data. Hyperparameter tuning is also done to find the best combination of parameters in support vector machines (SVM). Our results show that the SVM cross-validated ten times has a more accurate training result. Furthermore, the grid search on SVM cross-validated ten times had more accurate training model results and achieved 88% accuracy on the test data.

Klasifikasi Penyakit Diabetes Pada Imbalanced Class Dataset Menggunakan Algoritme Stacking

Diabetes is a disease that has the potential to cause death. Based on a report from the IDF (International Diabetes Federation), it was stated that in 2019 there were 463 million people in the world suffering from this disease. According to the Ministry of Health, Indonesia is a country that is included in the top 10 highest in the world by the number of people with diabetes. Machine learning models can be a solution for the early detection of diabetes based on history and available data. The majority of the research that has been done chiefly uses a single classifier. The single classifier model has a weakness when faced with class imbalance conditions in the dataset. Therefore, this study uses the Stacking Model for the classification and prediction process on the diabetes dataset. The goal is to improve the performance of a single classifier. In addition, the Stacking Model is expected to be one of the solutions for the classification of diabetes in the imbalanced class dataset. Based on two test experiments that have been carried out using two different datasets. The Stacking algorithm can produce an accuracy value of 89%, TPR value of 89%, TNR value of 85%, and G-Mean of 86.98% in the first dataset and can produce an accuracy value of 96%, TPR value of 96%, TNR value of 94%, and G-Mean of 94.99% in the second dataset. These results are better than single classifiers such as C4.5, K-NN, and SVM of the four indicators evaluated in both diabetes datasets. Thus, the proposed algorithm, namely Stacking (C4.5-SVM), can be a solution for classifying diabetes datasets with unbalanced class distribution conditions.

A Hybrid Feature Selection and Ensemble Approach to Identify Depressed Users in Online Social Media.

Depression has become one of the most common mental illnesses, and the widespread use of social media provides new ideas for detecting various mental illnesses. The purpose of this study is to use machine learning technology to detect users of depressive patients based on user-shared content and posting behaviors in social media. At present, the existing research mostly uses a single detection method, and the unbalanced class distribution often leads to a low recognition rate. In addition, a large number of irrelevant or redundant features in high-dimensional data sets interfere with the accuracy of recognition. To solve this problem, this paper proposes a hybrid feature selection and stacking ensemble strategy for depression user detection. First, recursive elimination method and extremely randomized trees method are used to calculate feature importance and mutual information value, calculate feature weight vector, and select the optimal feature subset according to the feature weight. Second, naive bayes, k-nearest neighbor, regularized logistic regression and support vector machine are used as base learners, and a simple logistic regression algorithm is used as a combination strategy to build a stacking model. Experimental results show that compared with other machine learning algorithms, the proposed hybrid method, which integrates feature selection and ensemble, has a higher accuracy of 90.27% in identifying online patients. We believe this study will help develop new methods to identify depressed people in social networks, providing guidance for future research.

Electric theft detection in advanced metering infrastructure using Jaya optimized combined Kernel‐Tree boosting classifier—A novel sequentially executed supervised machine learning approach

This paper presents a novel, sequentially executed supervised machine learning-based electric theft detection framework using a Jaya-optimized combined Kernel and Tree Boosting (KTBoost) classifier. It utilizes the intelligence of the XGBoost algorithm to estimate the missing values in the acquired dataset during the data pre-processing phase. An oversampling algorithm based on the Robust-SMOTE technique is utilized to avoid the unbalanced data class distribution issue. Afterward, with the aid of few very significant statistical, temporal, and spectral features extracted from the acquired kWh dataset, the complex underlying data patterns are comprehended to enhance the accuracy and detection rate of the classifier. For effectively classifying the consumers into “Honest” and “Fraudster,” the ensemble machine learning-based classifier KTBoost, with Jaya algorithm optimized hyperparameters, is utilized. Finally, the developed model is re-trained using a reduced set of highly important features to minimize the computational resources without compromising the performance of the developed model. The outcome of this study reveals that the proposed theft detection method achieves the highest accuracy (93.38%), precision (95%), and recall (93.18%) among all the studied methods, thus signifying its importance in the studied area of research.

Instance Switching-Based Contrastive Learning for Fine-Grained Airplane Detection

Detecting airplanes from high-resolution remote sensing images has a variety of applications. The characteristics of clear details, rich spatial and texture information of objects in high-resolution remote sensing images make it possible to identify different types of airplanes from backgrounds. However, airplanes usually exhibit slight inter-class discrepancy and unbalanced class distribution, which pose significant challenges to fine-grained detection of airplanes. In this paper, we propose the ISCL, an Instance Switching-based Contrastive Learning method for fine-grained airplane detection. Specifically, we introduce a Contrastive Learning-based Module (CLM) to widen the inter-class distance while narrowing the intra-class distance by optimizing feature space distribution with the InfoNCE <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">+</sup> loss, which is built on a serial head in a cascaded way. Then, we design a Refined Instance Switching (ReIS) module to alleviate the class imbalance problem. To take full advantage of the CLM and ReIS, we further introduce an optimization strategy which is an organic combination of the two modules to widen the distances of different airplane categories that are easily confused. In addition, we contribute a fine-grained attribute-assisted dataset, dubbed GF-RarePlanes Dataset (GRD), to help the detectors better learn the subtle differences between the airplanes. Extensive experiments on two datasets ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i.e</i> ., GF and FAIR1M) demonstrate that our proposed method can significantly improve the accuracy of fine-grained airplane detection under both HBB and OBB scenarios. Dataset and codes will be available at https://lanxin1011.github.io/ISCL/.

ÖNERİLEN KONVOLÜSYON SİNİR AĞI YAKLAŞIMI KULLANARAK ELMA YAPRAĞI HASTALIKLARININ SINIFLANDIRILMASI

It is difficult to constantly control apple trees in farmland. In case of a disease on tree leaves, the risk of disease transmission to other leaves is high. It is necessary to prevent further deterioration of the plant by performing automatic detection of the disease in the early period. If the disease detection is delayed, the planned production cannot be realized. It is too late if diseases are detected by a farmer or agronomist. In addition, as the agricultural lands grow, the number of experts needed increases accordingly. For these reasons, leaf images of apple trees are grouped into 4 different classes: apple peel, leaf rust, healthy apple and multiple disease states. In the proposed method, noise removal in the images, detection of the relevant area and histogram equalization on the YUV color space are performed. Due to the unbalanced class distribution in the data set used, data augmentation was applied for the minority classes with the SMOTE method. Afterwards, features are extracted using pre-trained network models DenseNet121, DenseNet201, InceptionResNetV2, InceptionV3, ResNet50V2. Extracted features were classified with a CNN-based method developed with an accuracy of 99%.

Securing BGP by Handling Dynamic Network Behavior and Unbalanced Datasets

The Border Gateway Protocol (BGP) provides crucial routing information for the Internet infrastructure. A problem with abnormal routing behavior affects the stability and connectivity of the global Internet. The biggest hurdles in detecting BGP attacks are extremely unbalanced data set category distribution and the dynamic nature of the network. This unbalanced class distribution and dynamic nature of the network results in the classifier's inferior performance. In this paper we proposed an efficient approach to properly managing these problems, the proposed approach tackles the unbalanced classification of datasets by turning the problem of binary classification into a problem of multiclass classification. This is achieved by splitting the majority-class samples evenly into multiple segments using Affinity Propagation, where the number of segments is chosen so that the number of samples in any segment closely matches the minority-class samples. Such sections of the dataset together with the minor class are then viewed as different classes and used to train the Extreme Learning Machine (ELM). The RIPE and BCNET datasets are used to evaluate the performance of the proposed technique. When no feature selection is used, the proposed technique improves the F1 score by 1.9% compared to state-of-the-art techniques. With the Fischer feature selection algorithm, the proposed algorithm achieved the highest F1 score of 76.3%, which was a 1.7% improvement over the compared ones. Additionally, the MIQ feature selection technique improves the accuracy by 3.5%. For the BCNET dataset, the proposed technique improves the F1 score by 1.8% for the Fisher feature selection technique. The experimental findings support the substantial improvement in performance from previous approaches by the new technique.

Towards Urban Scene Semantic Segmentation with Deep Learning from LiDAR Point Clouds: A Case Study in Baden-Württemberg, Germany

An accurate understanding of urban objects is critical for urban modeling, intelligent infrastructure planning and city management. The semantic segmentation of light detection and ranging (LiDAR) point clouds is a fundamental approach for urban scene analysis. Over the last years, several methods have been developed to segment urban furniture with point clouds. However, the traditional processing of large amounts of spatial data has become increasingly costly, both time-wise and financially. Recently, deep learning (DL) techniques have been increasingly used for 3D segmentation tasks. Yet, most of these deep neural networks (DNNs) were conducted on benchmarks. It is, therefore, arguable whether DL approaches can achieve the state-of-the-art performance of 3D point clouds segmentation in real-life scenarios. In this research, we apply an adapted DNN (ARandLA-Net) to directly process large-scale point clouds. In particular, we develop a new paradigm for training and validation, which presents a typical urban scene in central Europe (Munzingen, Freiburg, Baden-Württemberg, Germany). Our dataset consists of nearly 390 million dense points acquired by Mobile Laser Scanning (MLS), which has a rather larger quantity of sample points in comparison to existing datasets and includes meaningful object categories that are particular to applications for smart cities and urban planning. We further assess the DNN on our dataset and investigate a number of key challenges from varying aspects, such as data preparation strategies, the advantage of color information and the unbalanced class distribution in the real world. The final segmentation model achieved a mean Intersection-over-Union (mIoU) score of 54.4% and an overall accuracy score of 83.9%. Our experiments indicated that different data preparation strategies influenced the model performance. Additional RGB information yielded an approximately 4% higher mIoU score. Our results also demonstrate that the use of weighted cross-entropy with inverse square root frequency loss led to better segmentation performance than when other losses were considered.

Bi-TLLDA and CSSVM based fault diagnosis of vehicle on-board equipment for high speed railway

Vehicle on-board equipment (VOBE) is a significant component of the control system of high-speed railway train, the fault diagnosis of VOBE mainly depends on maintenance experience, which is inefficiency. The fault data of on-board equipment is described by natural language. Due to its unstructured, high-dimensional and unbalanced fault class distribution, it has become a challenge in fault diagnosis. In this paper, bilevel topic labeled latent Dirichlet allocation for extraction feature of fault text data is proposed. Firstly, label information is set according to prior knowledge of railway field. Then, local topics and global topics are defined for two level fault types of VOBE. The fault feature space generated by Gibbs sampling from local topics and global topics contains two level fault features, which are conducive to fault text classification. Finally, considering the imbalanced distribution of fault class, cost sensitive support vector machine based fault text classification is proposed. By using actual fault data of on-board equipment of China Railway Corporation, the accuracy, precision and F1-score are used as performance indicators to compare the proposed method with other fault diagnosis methods. The results show that the accuracy of the proposed method is 90.3%, which is about 2% higher than that of the suboptimal method, and the average recall, precision and F1-score of various fault classes are 77.9%, 91.8% and 83.4%, respectively, which outperforms other fault diagnosis methods.

A Weighted Discriminative Extreme Learning Machine Design for Lung Cancer Detection by an Electronic Nose System

This article presents a study on lung cancer detection based on electronic nose technology. The pattern recognition algorithm is extremely crucial for an electronic nose system, but the customary learning algorithms usually prefer the majority class for class imbalance learning due to the assumption of equal misclassification costs. To address this challenge, in this article, we propose a novel classification method named weighted discriminative extreme learning machine (WDELM) for lung cancer diagnosis. First, the WDELM assigns a different weight to each particular sample by using a flexible weighting strategy, which enables it to perform classification tasks with the unbalanced class distribution. Then, an alternating iterative algorithm is proposed to solve the convex objective function with a theoretical analysis presented. Finally, the effectiveness of the proposed method has been evaluated on lung cancer datasets and public datasets by comparing it with conventional methods. Experimental results have confirmed that the WDELM surpasses conventional methods and is competent for lung cancer diagnosis.