• ETRI Journal
• /
• v.39 no.5
• /
• pp.621-631
• /
• 2017

Abnormal samples are usually difficult to obtain in production systems, resulting in imbalanced training sample sets. Namely, the number of positive samples is far less than the number of negative samples. Traditional Support Vector Machine (SVM)-based anomaly detection algorithms perform poorly for highly imbalanced datasets: the learned classification hyperplane skews toward the positive samples, resulting in a high false-negative rate. This article proposes a new imbalanced SVM (termed ImSVM)-based anomaly detection algorithm, which assigns a different weight for each positive support vector in the decision function. ImSVM adjusts the learned classification hyperplane to make the decision function achieve a maximum GMean measure value on the dataset. The above problem is converted into an unconstrained optimization problem to search the optimal weight vector. Experiments are carried out on both Cloud datasets and Knowledge Discovery and Data Mining datasets to evaluate ImSVM. Highly imbalanced training sample sets are constructed. The experimental results show that ImSVM outperforms over-sampling techniques and several existing imbalanced SVM-based techniques.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.3
• /
• pp.591-609
• /
• 2024

In this study, preprocessings with all combinations were examined in terms of the effects on decreasing word number, shortening the duration of the process and the classification success in balanced and imbalanced datasets which were unbalanced in different ratios. The decreases in the word number and the processing time provided by preprocessings were interrelated. It was seen that more successful classifications were made with Turkish datasets and English datasets were affected more from the situation of whether the dataset is balanced or not. It was found out that the incorrect classifications, which are in the classes having few documents in highly imbalanced datasets, were made by assigning to the class close to the related class in terms of topic in Turkish datasets and to the class which have many documents in English datasets. In terms of average scores, the highest classification was obtained in Turkish datasets as follows: with not applying lowercase, applying stemming and removing stop words, and in English datasets as follows: with applying lowercase and stemming, removing stop words. Applying stemming was the most important preprocessing method which increases the success in Turkish datasets, whereas removing stop words in English datasets. The maximum scores revealed that feature selection, feature size and classifier are more effective than preprocessing in classification success. It was concluded that preprocessing is necessary for text classification because it shortens the processing time and can achieve high classification success, a preprocessing method does not have the same effect in all languages, and different preprocessing methods are more successful for different languages.

• International journal of advanced smart convergence
• /
• v.8 no.1
• /
• pp.75-81
• /
• 2019

Data imbalance problem is common and causes serious problem in machine learning process. Sampling is one of the effective methods for solving data imbalance problem. Over-sampling increases the number of instances, so when over-sampling is applied in imbalanced data, it is applied to minority instances. Under-sampling reduces instances, which usually is performed on majority data. We apply under-sampling and over-sampling to imbalanced data and generate sampled data sets. From the generated data sets from sampling and original data set, we construct a heterogeneous ensemble of classifiers. We apply five different algorithms to the heterogeneous ensemble. Experimental results on an intrusion detection dataset as an imbalanced datasets show that our approach shows effective results.

• The Journal of the Acoustical Society of Korea
• /
• v.42 no.4
• /
• pp.329-344
• /
• 2023

In this study, we comprehensively analyze the generalization performance of various deep learning-based active sonar target classifiers when applied to small and imbalanced active sonar datasets. To generate the active sonar datasets, we use data from two different oceanic experiments conducted at different times and ocean. Each sample in the active sonar datasets is a time-frequency domain image, which is extracted from audio signal of contact after the detection process. For the comprehensive analysis, we utilize 22 Convolutional Neural Networks (CNN) models. Two datasets are used as train/validation datasets and test datasets, alternatively. To calculate the variance in the output of the target classifiers, the train/validation/test datasets are repeated 10 times. Hyperparameters for training are optimized using Bayesian optimization. The results demonstrate that shallow CNN models show superior robustness and generalization performance compared to most of deep CNN models. The results from this paper can serve as a valuable reference for future research directions in deep learning-based active sonar target classification.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.41 no.2
• /
• pp.56-64
• /
• 2018

In recent years, imbalanced data is one of the most important and frequent issue for quality control in industrial field. As an example, defect rate has been drastically reduced thanks to highly developed technology and quality management, so that only few defective data can be obtained from production process. Therefore, quality classification should be performed under the condition that one class (defective dataset) is even smaller than the other class (good dataset). However, traditional multi-class classification methods are not appropriate to deal with such an imbalanced dataset, since they classify data from the difference between one class and the others that can hardly be found in imbalanced datasets. Thus, one-class classification that thoroughly learns patterns of target class is more suitable for imbalanced dataset since it only focuses on data in a target class. So far, several one-class classification methods such as one-class support vector machine, neural network and decision tree there have been suggested. One-class support vector machine and neural network can guarantee good classification rate, and decision tree can provide a set of rules that can be clearly interpreted. However, the classifiers obtained from the former two methods consist of complex mathematical functions and cannot be easily understood by users. In case of decision tree, the criterion for rule generation is ambiguous. Therefore, as an alternative, a new one-class classifier using hyper-rectangles was proposed, which performs precise classification compared to other methods and generates rules clearly understood by users as well. In this paper, we suggest an approach for improving the limitations of those previous one-class classification algorithms. Specifically, the suggested approach produces more improved one-class classifier using hyper-rectangles generated by using Gaussian function. The performance of the suggested algorithm is verified by a numerical experiment, which uses several datasets in UCI machine learning repository.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.27 no.6
• /
• pp.1385-1395
• /
• 2017

This paper aims to develop an IDS (Intrusion Detection System) that takes into account class imbalanced datasets. For this, we first built a set of training data sets from the Kyoto 2006+ dataset in which the amounts of normal data and abnormal (intrusion) data are not balanced. Then, we have run a number of tests to evaluate the effectiveness of machine learning techniques for detecting intrusions. Our evaluation results demonstrated that the Random Forest algorithm achieved the best performances.

• The Korean Journal of Applied Statistics
• /
• v.17 no.3
• /
• pp.445-457
• /
• 2004

Given class-imbalanced data in two-class classification problem, we often do over-sampling and/or under-sampling of training data to make it balanced. We investigate the validity of such practice. Also we study the effect of such sampling practice on boosting of classification trees. Through experiments on twelve real datasets it is observed that keeping the natural distribution of training data is the best way if you plan to apply boosting methods to class-imbalanced data.

• Information Systems Review
• /
• v.24 no.4
• /
• pp.23-40
• /
• 2022

Although Support Vector Machine(SVM) has been used in various fields such as bankruptcy prediction model, the hyperplane learned by SVM in class imbalance problem can be severely skewed toward minority class and has a negative impact on performance because the area of majority class is expanded while the area of minority class is invaded. This study proposed optimized uneven margin SVM(OPT-UMSVM) combining threshold moving or post scaling method with UMSVM to cope with the limitation of the traditional even margin SVM(EMSVM) in class imbalance problem. OPT-UMSVM readjusted the skewed hyperplane to the majority class and had better generation ability than EMSVM improving the sensitivity of minority class and calculating the optimized performance. To validate OPT-UMSVM, 10-fold cross validations were performed on five sub-datasets with different imbalance ratio values. Empirical results showed two main findings. First, UMSVM had a weak effect on improving the performance of EMSVM in balanced datasets, but it greatly outperformed EMSVM in severely imbalanced datasets. Second, compared to EMSVM and conventional UMSVM, OPT-UMSVM had better performance in both balanced and imbalanced datasets and showed a significant difference performance especially in severely imbalanced datasets.

• International Journal of Internet, Broadcasting and Communication
• /
• v.11 no.4
• /
• pp.37-42
• /
• 2019

In this paper, we explore the details of three classic data augmentation methods and two generative model based oversampling methods. The three classic data augmentation methods are random sampling (RANDOM), Synthetic Minority Over-sampling Technique (SMOTE), and Adaptive Synthetic Sampling (ADASYN). The two generative model based oversampling methods are Conditional Generative Adversarial Network (CGAN) and Wasserstein Generative Adversarial Network (WGAN). In imbalanced data, the whole instances are divided into majority class and minority class, where majority class occupies most of the instances in the training set and minority class only includes a few instances. Generative models have their own advantages when they are used to generate more plausible samples referring to the distribution of the minority class. We also adopt CGAN to compare the data augmentation performance with other methods. The experimental results show that WGAN-based oversampling technique is more stable than other approaches (RANDOM, SMOTE, ADASYN and CGAN) even with the very limited training datasets. However, when the imbalanced ratio is too small, generative model based approaches cannot achieve satisfying performance than the conventional data augmentation techniques. These results suggest us one of future research directions.

• KIPS Transactions on Software and Data Engineering
• /
• v.12 no.7
• /
• pp.295-302
• /
• 2023

Software requirements written in natural language may have different meanings from the stakeholders' viewpoint. When designing an architecture based on quality attributes, it is necessary to accurately classify quality attribute requirements because the efficient design is possible only when appropriate architectural tactics for each quality attribute are selected. As a result, although many natural language processing models have been studied for the classification of requirements, which is a high-cost task, few topics improve classification performance with the imbalanced quality attribute datasets. In this study, we first show that the classification model can automatically classify the Korean requirement dataset through experiments. Based on these results, we explain that data augmentation through EDA(Easy Data Augmentation) techniques and undersampling strategies can improve the imbalance of quality attribute datasets, and show that they are effective in classifying requirements. The results improved by 5.24%p on F1-score, indicating that handling imbalanced data helps classify Korean requirements of classification models. Furthermore, detailed experiments of EDA illustrate operations that help improve classification performance.