• Journal of the Korea Convergence Society
• /
• 제12권1호
• /
• pp.49-55
• /
• 2021

Recently, with the development of database, it is possible to store a lot of data generated in finance, security, and networks. These data are being analyzed through classifiers based on machine learning. The main problem at this time is data imbalance. When we train imbalanced data, it may happen that classification accuracy is degraded due to over-fitting with majority class data. To overcome the problem of data imbalance, oversampling strategy that increases the quantity of data of minority class data is widely used. It requires to tuning process about suitable method and parameters for data distribution. To improve the process, In this study, we propose a strategy to explore and optimize oversampling combinations and ratio based on various methods such as synthetic minority oversampling technique and generative adversarial networks through genetic algorithms. After sampling credit card fraud detection which is a representative case of data imbalance, with the proposed strategy and single oversampling strategies, we compare the performance of trained classifiers with each data. As a result, a strategy that is optimized by exploring for ratio of each method with genetic algorithms was superior to previous strategies.

• Earthquakes and Structures
• /
• 제27권3호
• /
• pp.177-189
• /
• 2024

Accurately predicting the failure modes of reinforced concrete (RC) columns is essential for structural design and assessment. In this study, the challenges of imbalanced datasets and complex feature selection in machine learning (ML) methods were addressed through an optimized ML approach. By combining feature selection and oversampling techniques, the prediction of seismic failure modes in rectangular RC columns was improved. Two feature selection methods were used to identify six input parameters. To tackle class imbalance, the Borderline-SMOTE1 algorithm was employed, enhancing the learning capabilities of the models for minority classes. Eight ML algorithms were trained and fine-tuned using k-fold shuffle split cross-validation and grid search. The results showed that the artificial neural network model achieved 96.77% accuracy, while k-nearest neighbor, support vector machine, and random forest models each achieved 95.16% accuracy. The balanced dataset led to significant improvements, particularly in predicting the flexure-shear failure mode, with accuracy increasing by 6%, recall by 8%, and F1 scores by 7%. The use of the Borderline-SMOTE1 algorithm significantly improved the recognition of samples at failure mode boundaries, enhancing the classification performance of models like k-nearest neighbor and decision tree, which are highly sensitive to data distribution and decision boundaries. This method effectively addressed class imbalance and selected relevant features without requiring complex simulations like traditional methods, proving applicable for discerning failure modes in various concrete members under seismic action.

• Ecology and Resilient Infrastructure
• /
• 제10권4호
• /
• pp.107-115
• /
• 2023

High turbidity in source water can have adverse effects on water treatment plant operations and aquatic ecosystems, necessitating turbidity management. Consequently, research aimed at predicting river turbidity continues. This study developed a multi-class classification model for prediction of turbidity using LightGBM (Light Gradient Boosting Machine), a representative ensemble machine learning algorithm. The model utilized data that was classified into four classes ranging from 1 to 4 based on turbidity, from low to high. The number of input data points used for analysis varied among classes, with 945, 763, 95, and 25 data points for classes 1 to 4, respectively. The developed model exhibited precisions of 0.85, 0.71, 0.26, and 0.30, as well as recalls of 0.82, 0.76, 0.19, and 0.60 for classes 1 to 4, respectively. The model tended to perform less effectively in the minority classes due to the limited data available for these classes. To address data imbalance, the SMOTE (Synthetic Minority Over-sampling Technique) algorithm was applied, resulting in improved model performance. For classes 1 to 4, the Precision and Recall of the improved model were 0.88, 0.71, 0.26, 0.25 and 0.79, 0.76, 0.38, 0.60, respectively. This demonstrated that alleviating data imbalance led to a significant enhancement in Recall of the model. Furthermore, to analyze the impact of differences in input data composition addressing the input data imbalance, input data was constructed with various ratios for each class, and the model performances were compared. The results indicate that an appropriate composition ratio for model input data improves the performance of the machine learning model.

• Journal of Intelligence and Information Systems
• /
• 제15권3호
• /
• pp.1-15
• /
• 2009

In a classification problem, data imbalance occurs when the number of instances in one class greatly outnumbers the number of instances in the other class. Such data sets often cause a default classifier to be built due to skewed boundary and thus the reduction in the classification accuracy of such a classifier. This paper proposes a Geometric Mean-based Boosting (GM-Boost) to resolve the problem of data imbalance. Since GM-Boost introduces the notion of geometric mean, it can perform learning process considering both majority and minority sides, and reinforce the learning on misclassified data. An empirical study with bankruptcy prediction on Korea companies shows that GM-Boost has the higher classification accuracy than previous methods including Under-sampling, Over-Sampling, and AdaBoost, used in imbalanced data and robust learning performance regardless of the degree of data imbalance.

• Journal of the Korea Convergence Society
• /
• 제12권3호
• /
• pp.17-25
• /
• 2021

When a person wears clothing, it is important to configure an outfit appropriate to the intended occasion. Therefore, T.P.O(Time, Place, Occasion) of the outfit is considered in various fashion recommendation systems based on artificial intelligence. However, there are few studies that directly infer the T.P.O from outfit images, as the nature of the problem causes multi-label and class imbalance problems, which makes model training challenging. Therefore, in this study, we propose a model that can infer the T.P.O of outfit images by employing a label-distribution-aware margin(LDAM) loss function. Datasets for the model training and evaluation were collected from fashion shopping malls. As a result of measuring performance, it was confirmed that the proposed model showed balanced performance in all T.P.O classes compared to baselines.

• Smart Media Journal
• /
• 제8권2호
• /
• pp.46-57
• /
• 2019

In recent research, deep learning-based methods have achieved state-of-the-art performance in various computer vision tasks. However, these methods are commonly supervised, and require huge amounts of annotated data to train. Acquisition of data demands an additional costly effort, particularly for the tasks where it becomes challenging to obtain large amounts of data considering the time constraints and the requirement of professional human diligence. In this paper, we present a data level synthetic sampling solution to learn from small and imbalanced data sets using Generative Adversarial Networks (GANs). The reason for using GANs are the challenges posed in various fields to manage with the small datasets and fluctuating amounts of samples per class. As a result, we present an approach that can improve learning with respect to data distributions, reducing the partiality introduced by class imbalance and hence shifting the classification decision boundary towards more accurate results. Our novel method is demonstrated on a small dataset of 2789 tomato plant disease images, highly corrupted with class imbalance in 9 disease categories. Moreover, we evaluate our results in terms of different metrics and compare the quality of these results for distinct classes.

• Journal of Industrial Convergence
• /
• 제22권1호
• /
• pp.177-183
• /
• 2024

The study investigates the impact of data quality on the performance of artificial intelligence (AI). To this end, the impact of labeling error levels on the performance of artificial intelligence was compared and analyzed through simulation, taking into account the similarity of data features and the imbalance of class composition. As a result, data with high similarity between characteristic variables were found to be more sensitive to labeling accuracy than data with low similarity between characteristic variables. It was observed that artificial intelligence accuracy tended to decrease rapidly as class imbalance increased. This will serve as the fundamental data for evaluating the quality criteria and conducting related research on artificial intelligence learning data.

• The Korean Journal of Applied Statistics
• /
• 제27권3호
• /
• pp.357-371
• /
• 2014

There are many studies related to imbalanced data in which the class distribution is highly skewed. To address the problem of imbalanced data, previous studies deal with resampling techniques which correct the skewness of the class distribution in each sampled subset by using under-sampling, over-sampling or hybrid-sampling such as SMOTE. Ensemble methods have also alleviated the problem of class imbalanced data. In this paper, we compare around a dozen algorithms that combine the ensemble methods and resampling techniques based on simulated data sets generated by the Backbone model, which can handle the imbalance rate. The results on various real imbalanced data sets are also presented to compare the effectiveness of algorithms. As a result, we highly recommend the resampling technique combining ensemble methods for imbalanced data in which the proportion of the minority class is less than 10%. We also find that each ensemble method has a well-matched sampling technique. The algorithms which combine bagging or random forest ensembles with random undersampling tend to perform well; however, the boosting ensemble appears to perform better with over-sampling. All ensemble methods combined with SMOTE outperform in most situations.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• 제48권11호
• /
• pp.861-871
• /
• 2020

Raising the speed of the momentum wheel in the CMG increases the unintended force and torque caused by mass imbalance. This unintended force and torque should be minimized to get the better quality of satellite SAR image because they lead to the vibration of the output image. This paper shows the works on compensating the static imbalance and couple mass imbalance in the CMG wheel. First, the force and torque at the center of mass generated by the mass imbalance were predicted through M&S analysis. Second, the force and torque were estimated similarly through the M&S analysis when the measurement point was moved from the rotation center. Third, the measurement configuration for the force and torque by the mass imbalance was described. Fourth, the change of the force and torque by adding the specified mass to the momentum wheel was observed after comparing the measurements with the results of the M&S. And finally, the effect of the compensation was analyzed by comparing the force and torque before and after the correction while 24Nm class CMG was running in the standby mode.

• The Korean Journal of Applied Statistics
• /
• 제32권3호
• /
• pp.349-374
• /
• 2019

A class imbalance problem arises when one class outnumbers the other class by a large proportion in binary data. Studies such as transforming the learning data have been conducted to solve this imbalance problem. In this study, we compared resampling methods among methods to deal with an imbalance in the classification problem. We sought to find a way to more effectively detect the minority class in the data. Through simulation, a total of 20 methods of over-sampling, under-sampling, and combined method of over- and under-sampling were compared. The logistic regression, support vector machine, and random forest models, which are commonly used in classification problems, were used as classifiers. The simulation results showed that the random under sampling (RUS) method had the highest sensitivity with an accuracy over 0.5. The next most sensitive method was an over-sampling adaptive synthetic sampling approach. This revealed that the RUS method was suitable for finding minority class values. The results of applying to some real data sets were similar to those of the simulation.