• International journal of advanced smart convergence
• /
• 제12권1호
• /
• pp.53-58
• /
• 2023

With the fast development of artificial intelligence day by day, users are demanding explanations about the results of algorithms and want to know what parameters influence the results. In this paper, we propose a model for bankruptcy prediction with interpretability using the SHAP framework. SHAP (SHAPley Additive exPlanations) is framework that gives a visualized result that can be used for explanation and interpretation of machine learning models. As a result, we can describe which features are important for the result of our deep learning model. SHAP framework Force plot result gives us top features which are mainly reflecting overall model score. Even though Fully Connected Neural Networks are a "black box" model, Shapley values help us to alleviate the "black box" problem. FCNNs perform well with complex dataset with more than 60 financial ratios. Combined with SHAP framework, we create an effective model with understandable interpretation. Bankruptcy is a rare event, then we avoid imbalanced dataset problem with the help of SMOTE. SMOTE is one of the oversampling technique that resulting synthetic samples are generated for the minority class. It uses K-nearest neighbors algorithm for line connecting method in order to producing examples. We expect our model results assist financial analysts who are interested in forecasting bankruptcy prediction of companies in detail.

• 한국지구과학회지
• /
• 제43권4호
• /
• pp.511-519
• /
• 2022

Given the significant social and economic impact caused by heat waves, there is a pressing need to predict them with high accuracy and reliability. In this study, we analyzed the real-time forecast data from six models constituting the Subseasonal-to-Seasonal (S2S) prediction project, to elucidate the key mechanisms contributing to the prediction of the recent record-breaking Korean heat wave event in 2018. Weekly anomalies were first obtained by subtracting the 2017-2020 mean values for both S2S model simulations and observations. By comparing four Korean heat-wave-related indices from S2S models to the observed data, we aimed to identify key climate processes affecting prediction accuracy. The results showed that superior performance at predicting the 2018 Korean heat wave was achieved when the model showed better prediction performance for the anomalous anticyclonic activity in the upper troposphere of Eastern Europe and the cyclonic circulation over the Western North Pacific (WNP) region compared to the observed data. Furthermore, the development of upper-tropospheric anticyclones in Eastern Europe was closely related to global warming and the occurrence of La Niña events. The anomalous cyclonic flow in the WNP region coincided with enhancements in Madden-Julian oscillation phases 4-6. Our results indicate that, for the accurate prediction of heat waves, such as the 2018 Korean heat wave, it is imperative for the S2S models to realistically reproduce the variabilities over the Eastern Europe and WNP regions.

• 한국대기환경학회지
• /
• 제19권2호
• /
• pp.191-204
• /
• 2003

A one dimensional time dependent acid rain model considering size distribution of aerosols and hydrometeors is developed to predict observed chemical and physical properties of precipitation. Temporal variations of anions and cations observed are predicted fairly well with acid rain model simulations. It is found that aerosol depletion rates are highly dependent on aerosol sizes under the assumption of Marshall - Palmer raindrop size distribution. Also, the aerosol depletion during the initial rain event largely influences on ion concentrations in rainwaters.

• 천문학회보
• /
• 제33권2호
• /
• pp.32.2-32.2
• /
• 2008

• International Journal of Automotive Technology
• /
• 제3권4호
• /
• pp.129-135
• /
• 2002

Durability estimation of a prototype vehicle has traditionally relied heavily on accelerated durability tests using predefined proving grounds or rig tests using a road simulator. By use of those tests, it is very difficult to predict durability failures in actual service environments. This motivated the development of an integrated CAE (Computer Aided Engineering) methodology for the durability estimation of a prototype vehicle in actual service environments. Since expensive computational costs such as computation time and hardware resources are required for a full vehicle simulation in those environments with a very long span of event time, the conventional CAE methodologies have little feasibility. An efficient computational methodology for durability estimations is applied with theoretical developments. The effectiveness of the proposed methodology is shown by the comparison of results of the typical actual service environment such as the city mode with those of the typical accelerated durability test over the Belgian road.

• Journal of Information Technology Applications and Management
• /
• 제22권3호
• /
• pp.1-18
• /
• 2015

Financial markets are characterized by large numbers of complex and interacting factors which are ill-understood and frequently difficult to measure. Mathematical models developed in finance are precise formulations of theories of how these factors interact to produce the market value of financial asset. While these models are quite good at predicting these market values, because these forces and their interactions are not precisely understood, the model value nevertheless deviates to some extent from the observable market value. In this paper we propose a framework for augmenting the predictive capabilities of mathematical model with a learning component which is primed with an initial set of historical data and then adjusts its behavior after the event of prediction.

• Industrial Engineering and Management Systems
• /
• 제13권4호
• /
• pp.449-453
• /
• 2014

Chronic diseases including hypertension and its complications are major sources causing the national medical expenditures to increase. We aim to predict the risk of hypertension complications for hypertension patients, using the sample national healthcare database established by Korean National Health Insurance Corporation. We apply classification techniques, such as logistic regression, linear discriminant analysis, and classification and regression tree to predict the hypertension complication onset event for each patient. The performance of these three methods is compared in terms of accuracy, sensitivity and specificity. The result shows that these methods seem to perform similarly although the logistic regression performs marginally better than the others.

• Journal of the Korean Data and Information Science Society
• /
• 제22권3호
• /
• pp.605-612
• /
• 2011

In multi-center randomized clinical trial the treatment eect may be changed over centers. It is thus important to investigate the heterogeneity in treatment eect between centers. For this, uncorrelated random-eect models assuming independence between random-eect terms have been often used, which may be a strong assumption. In this paper we propose a correlated frailty modelling approach of investigating such heterogeneity using the hierarchical-likelihood method when the outcome is time-to-event. In particular, we show how to construct a proper prediction interval for frailty, which explores graphically the potential heterogeneity for a treatment-by-center interaction term. The proposed method is illustrated via numerical studies based on data from the design of a multi-center clinical trial.

• 한국전산유체공학회:학술대회논문집
• /
• 한국전산유체공학회 2006년도 추계 학술대회논문집
• /
• pp.41-46
• /
• 2006

When new weapons are introduced, the target points estimation is one of the important objectives in the flight test as well as the safe separation. The prediction methods help to design the flight test schedule. However, the incremental aerodynamic coefficients in the aircraft flow field so-called BSE are difficult to predict. Generally, the semiempirical methods such as the grid methods, IFM and Flow TGP using database are used for estimation of BSE. However, these methods are quasi-steady methods using static aerodynamic loads. Nowadays the time-accurate CFD method is often used to predict the store separation event. In the current process, the incremental aerodynamic coefficients in BSE regime are calculated directly, and the elimination of delta coefficients is checked simultaneously. This stage can be used for the initial condition of Flow TGP with freestream database. Two dimensional supersonic and subsonic store separation problems have been simulated and incremental coefficients are calculated. The results show the time when the store gets out of BSE region.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2003년도 학술회의 논문집 정보 및 제어부문 A
• /
• pp.283-286
• /
• 2003

본 연구는 기존의 고장진단 기법들을 토대로 주어진 자동화 시스템에 실제 적용이 가능한 고장예측 알고리즘을 제시한다. 고장예측은 시스템이 운용되는 도중에 제한된 정보와 컴퓨터 자원을 이용하여 수행되어야 하므로 실시간 적용을 위하여 2단계로 구분하여 수행된다. 첫 번째는 실시간 고장예측 단계로서 시스템 운용 중에 시스템의 고장 징후를 탐지하는 역할을 하며, 두 번째는 오프라인 고장예측 단계로서 실시간으로 고장 징후가 탐지되면 시스템의 작동을 멈춘 후 고장의 징후를 분류하고 식별하는 역할을 수행한다 원활한 고장예측 알고리즘을 도출하기 위해 자동화 시스템의 이산사건 모델과 연속시간 모델을 수립하였으며, 이들을 통합한 공정모델에 대하여 하이브리드 시뮬레이션 환경을 구축하였다. 제안된 기법은 자동화 시스템의 공정모델에 기구부, 모터부에 대한 고장모델을 부가하여 컴퓨터 시뮬레이션을 통하여 타당성을 검증하였다.