과도한 조류 발생은 수생태계 교란과 수질 악화를 초래하는 대표적인 환경 문제로, 효과적인 관리와 대응을 위해 정확한 예측이 필요하다. 우리나라는 사계절의 기후 특성이 뚜렷하며, 수온이 상승하는 하절기에 조류 발생이 집중되는 경향을 보인다. 이에 따라 실시간 모니터링 자료는 대부분 저농도 상태가 유지되어 데이터 불균형 문제가 발생한다. 본 연구에서는 chlorophyll-a 농도를 기준으로 하천 현장의 조류 발생 수준을 Class 1 (Chl-a ≤ 10 ㎍/L), Class 2 (10 < Chl-a ≤ 50 ㎍/L), Class 3 (Chl-a > 50 ㎍/L)와 같이 3개의 class로 구분하고, 대표적인 앙상블 머신러닝 모형인 extreme gradient boosting (XGB) 알고리즘을 이용하여 조류 발생 수준을 예측하는 분류 모형을 구축하였다. 데이터 불균형 해소를 위해 생성형 인공지능 기반 알고리즘인 conditional generative adversarial network (CGAN)과 전통적인 데이터 보강 알고리즘인 synthetic minority over-sampling technique (SMOTE), 그리고 딥러닝 기반 기법인 autoencoder (AE)를 활용한 3가지 데이터 보강 알고리즘을 활용하여 데이터의 불균형을 개선한 자료를 생성하고 이를 XGB 모형에 적용하여 성능 변화를 비교하였다. 분석 결과 macro average 기준으로 원본 데이터를 사용한 모형의 recall은 0.606이었으나 SMOTE, AE 및 CGAN의 recall은 각각 0.666, 0.682, 0.720으로 크게 개선되었고, F1 score도 데이터 불균형 해소를 통해 약 7–13%의 성능이 향상되는 등 전체적으로 데이터 불균형 해소로 모형의 성능이 향상되었으며 CGAN이 가장 우수한 성능 개선 효과를 보이는 것으로 나타냈다. 본 연구의 결과를 통해 데이터 불균형 해소를 통한 머신러닝 모형 성능 개선 가능성을 확인하였다.

This study aims to improve the interpretability and transparency of forecasting results by applying an explainable AI technique to corporate default prediction models. In particular, the research addresses the challenges of data imbalance and the economic cost asymmetry of forecast errors. To tackle these issues, predictive performance was analyzed using the SMOTE-ENN imbalance sampling technique and a cost-sensitive learning approach. The main findings of the study are as follows. First, the four machine learning models used in this study (Logistic Regression, Random Forest, XGBoost, and CatBoost) produced significantly different evaluation results depending on the degree of asymmetry in forecast error costs between imbalance classes and the performance metrics applied. Second, XGBoost and CatBoost showed good predictive performance when considering variations in prediction cost asymmetry and diverse evaluation metrics. In particular, XGBoost showed the smallest gap between the actual default rate and the default judgment rate, highlighting its robustness in handling class imbalance and prediction cost asymmetry. Third, SHAP analysis revealed that total assets, net income to total assets, operating income to total assets, financial liability to total assets, and the retained earnings ratio were the most influential factors in predicting defaults. The significance of this study lies in its comprehensive evaluation of predictive performance of various ML models under class imbalance and cost asymmetry in forecast errors. Additionally, it demonstrates how explainable AI techniques can enhance the transparency and reliability of corporate default prediction models.

This study investigates using Conditional Tabular Generative Adversarial Networks (CT-GAN) to generate synthetic data for turnover prediction in large employment datasets. The effectiveness of CT-GAN is compared with Adaptive Synthetic Sampling (ADASYN), Synthetic Minority Over-sampling Technique (SMOTE), and Random Oversampling (ROS) using Logistic Regression (LR), Linear Discriminant Analysis (LDA), Random Forest (RF), and Extreme Learning Machines (ELM), evaluated with AUC and F1-scores. Results show that GAN-based techniques, especially CT-GAN, outperform traditional methods in addressing data imbalance, highlighting the need for advanced oversampling methods to improve classification accuracy in imbalanced datasets.

The injection molding process is a process in which thermoplastic resin is heated and made into a fluid state, injected under pressure into the cavity of a mold, and then cooled in the mold to produce a product identical to the shape of the cavity of the mold. It is a process that enables mass production and complex shapes, and various factors such as resin temperature, mold temperature, injection speed, and pressure affect product quality. In the data collected at the manufacturing site, there is a lot of data related to good products, but there is little data related to defective products, resulting in serious data imbalance. In order to efficiently solve this data imbalance, undersampling, oversampling, and composite sampling are usally applied. In this study, oversampling techniques such as random oversampling (ROS), minority class oversampling (SMOTE), ADASYN(Adaptive Synthetic Sampling), etc., which amplify data of the minority class by the majority class, and complex sampling using both undersampling and oversampling, are applied. For composite sampling, SMOTE+ENN and SMOTE+Tomek were used. Artificial neural network techniques is used to predict product quality. Especially, MLP and RNN are applied as artificial neural network techniques, and optimization of various parameters for MLP and RNN is required. In this study, we proposed an SA technique that optimizes the choice of the sampling method, the ratio of minority classes for sampling method, the batch size and the number of hidden layer units for parameters of MLP and RNN. The existing sampling methods and the proposed SA method were compared using accuracy, precision, recall, and F1 Score to prove the superiority of the proposed method.