• 한국인터넷방송통신학회논문지
• /
• 제22권4호
• /
• pp.111-117
• /
• 2022

수십년간 매우 많은 소프트웨어 결함 예측 모델에 관한 연구들이 수행되었으며, 그들 중 기계학습 기법을 사용한 모델들이 가장 좋은 성능을 보였다. 딥러닝 기법은 기계학습 분야에서 가장 각광받는 기술이 되었지만 결함 예측 모델의 분류기로 사용된 연구는 거의 없었다. 몇몇 연구들은 모델의 입력 소스나 구문 데이터로부터 시맨틱 정보를 얻어내는데 딥러닝을 사용하였다. 본 논문은 3개 이상의 은닉층을 갖는 MLP를 이용하여 모델 구조와 하이퍼 파라미터를 변경하여 여러 모델들을 제작하였다. 모델 평가 실험 결과 MLP 기반 딥러닝 모델들은 기존 결함 예측 모델들과 Accuracy는 비슷한 성능을 보였으나 AUC는 유의미하게 더 우수한 성능을 보였다. 또한 또다른 딥러닝 모델인 CNN 모델보다도 더 나은 성능을 보였다.

• Journal of Korea Artificial Intelligence Association
• /
• 제1권1호
• /
• pp.11-16
• /
• 2023

Accurate hospital case modeling and prediction are crucial for efficient healthcare. In this study, we demonstrate the implementation of regression analysis methods in machine learning systems utilizing mathematical statics and machine learning techniques. The developed machine learning model includes Bayesian linear, artificial neural network, decision tree, decision forest, and linear regression analysis models. Through the application of these algorithms, corresponding regression models were constructed and analyzed. The results suggest the potential of leveraging machine learning systems for medical research. The experiment aimed to create an Azure Machine Learning Studio tool for the speedy evaluation of multiple regression models. The tool faciliates the comparision of 5 types of regression models in a unified experiment and presents assessment results with performance metrics. Evaluation of regression machine learning models highlighted the advantages of boosted decision tree regression, and decision forest regression in hospital case prediction. These findings could lay the groundwork for the deliberate development of new directions in medical data processing and decision making. Furthermore, potential avenues for future research may include exploring methods such as clustering, classification, and anomaly detection in healthcare systems.

• Genomics & Informatics
• /
• 제14권4호
• /
• pp.149-159
• /
• 2016

With the success of the genome-wide association studies (GWASs), many candidate loci for complex human diseases have been reported in the GWAS catalog. Recently, many disease prediction models based on penalized regression or statistical learning methods were proposed using candidate causal variants from significant single-nucleotide polymorphisms of GWASs. However, there have been only a few systematic studies comparing existing methods. In this study, we first constructed risk prediction models, such as stepwise linear regression (SLR), least absolute shrinkage and selection operator (LASSO), and Elastic-Net (EN), using a GWAS chip and GWAS catalog. We then compared the prediction accuracy by calculating the mean square error (MSE) value on data from the Korea Association Resource (KARE) with body mass index. Our results show that SLR provides a smaller MSE value than the other methods, while the numbers of selected variables in each model were similar.

• 한국전산유체공학회지
• /
• 제11권3호
• /
• pp.59-63
• /
• 2006

Accurate prediction of a supersonic missile base drag continues to defy even well-rounded CFD codes. In an effort to address the accuracy and predictability of the base drags, the influence of grid system and competitive turbulence models on the base drag is analyzed. Characteristics of some turbulence models is reviewed through incompressible turbulent flow over a flat plate, and performance for the base drag prediction of several turbulence models such as Baldwin-Loman(B-L), Spalart-Allmaras(S-A), k-$\varepsilon$, k-$\omega$ model is assessed. When compressibility correction is injected into the S-A model, prediction accuracy of the base drag is enhanced. The NSWC wind tunnel test data are utilized for comparison of CFD and semi-empirical codes on the accuracy of base drag predictability: they are about equal, but CFD tends to perform better. It is also found that, as angle of attack of a missile with control fins increases, even the best CFD analysis tool we have lacks the accuracy needed for the base drag prediction.

• 대한기계학회논문집B
• /
• 제21권6호
• /
• pp.804-812
• /
• 1997

The present study has tested most of loss models previously published in the open literature and found an optimum set of empirical loss models for a reliable performance prediction of centrifugal compressors. In order to improve the prediction of efficiency curves, this paper recommends a modified parasitic loss model. Predicted performance curves by the proposed optimum set agree fairly well with experimental data for a variety of centrifugal compressors. The prediction method developed through this study can serve as a tool for preliminary design and assist the understanding of the operational characteristics of general purpose centrifugal compressors.

• 응용통계연구
• /
• 제17권2호
• /
• pp.219-228
• /
• 2004

오차항의 분포가 정규분포에 따르지 않는 비선형 시계열인 ARCH모형의 예측구간을 설정하는데 붓스트랩 방법과 근사적 방법간의 포함비율에 대한 정확성을 비교한다. 이 때 모형에서 모수를 추정하는 방법으로서는 분포에 대한 가정을 필요로 하지 않는 quasi-score 추정함수를 이용한 추정 법과 로버스트 추정 함수인 M quasi-score 추정 함수를 이용한 추정법을 사용한다. 추정된 모수를 이용하여 예측구간의 정확성을 비교하고 마지막으로 소비자 물가지수 자료를 이용하여 실제 예측구간을 구하는데 적용한다.

• 한국정밀공학회:학술대회논문집
• /
• 한국정밀공학회 2006년도 춘계학술대회 논문집
• /
• pp.351-352
• /
• 2006

There are many methodologies fur doing analysis of system's reliability in early design stage. Among the methods, PRISM is, as compared to MIL-HDBK-217, a newly developed technology but not easy to use. Because PRISM provides models that predict a part failure rate and field database, called EPRD and NPRD that can be combined with prediction models. This paper presents some capabilities of the prediction models in PRISM and usability of EPRD and NPRD database in system level reliability prediction.

• Journal of Information Processing Systems
• /
• 제10권2호
• /
• pp.314-333
• /
• 2014

When analyzing default predictions in real estate companies, the number of non-defaulted cases always greatly exceeds the defaulted ones, which creates the two-class imbalance problem. This lowers the ability of prediction models to distinguish the default sample. In order to avoid this sample selection bias and to improve the prediction model, this paper applies a minority sample generation approach to create new minority samples. The logistic regression, support vector machine (SVM) classification, and neural network (NN) classification use an imbalanced dataset. They were used as benchmarks with a single prediction model that used a balanced dataset corrected by the minority samples generation approach. Instead of using prediction-oriented tests and the overall accuracy, the true positive rate (TPR), the true negative rate (TNR), G-mean, and F-score are used to measure the performance of default prediction models for imbalanced dataset. In this paper, we describe an empirical experiment that used a sampling of 14 default and 315 non-default listed real estate companies in China and report that most results using single prediction models with a balanced dataset generated better results than an imbalanced dataset.

• Computers and Concrete
• /
• 제29권 6호
• /
• pp.433-444
• /
• 2022

Concrete's compressive strength is widely studied in order to understand many qualities and the grade of the concrete mixture. Conventional civil engineering tests involve time and resources consuming laboratory operations which results in the deterioration of concrete samples. Proposing efficient non-destructive models for the prediction of concrete compressive strength will certainly yield advancements in concrete studies. In this study, the efficiency of using radial basis function neural network (RBFNN) which is not common in this field, is studied for the concrete compressive strength prediction. Complementary studies with back propagation neural network (BPNN), which is commonly used in this field, have also been carried out in order to verify the efficiency of RBFNN for compressive strength prediction. A total of 13 input parameters, including novel ones such as cement's and fly ash's compositional information, have been employed in the prediction models with RBFNN and BPNN since all these parameters are known to influence concrete strength. Three different train: test ratios were tested with both models, while different hidden neurons, epochs, and spread values were introduced to determine the optimum parameters for yielding the best prediction results. Prediction results obtained by RBFNN are observed to yield satisfactory high correlation coefficients and satisfactory low mean square error values when compared to the results in the previous studies, indicating the efficiency of the proposed model.

• 대기
• /
• 제28권3호
• /
• pp.291-303
• /
• 2018

In this study, a new consensus technique for predicting tropical cyclone (TC) intensity in the western North Pacific was developed. The most important feature of the present consensus model is to select and combine the guidance numerical models with the best performance in the previous years based on various evaluation criteria and averaging methods. Specifically, the performance of the guidance models was evaluated using both the mean absolute error and the correlation coefficient for each forecast lead time, and the number of the numerical models used for the consensus model was not fixed. In averaging multiple models, both simple and weighted methods are used. These approaches are important because that the performance of the available guidance models differs according to forecast lead time and is changing every year. In particular, this study develops both a multi-consensus model (M-CON), which constructs the best consensus models with the lowest error for each forecast lead time, and a single best consensus model (S-CON) having the lowest 72-hour cumulative mean error, through on training process. The evaluation results of the selected consensus models for the training and forecast periods reveal that the M-CON and S-CON outperform the individual best-performance guidance models. In particular, the M-CON showed the best overall performance, having advantages in the early stages of prediction. This study finally suggests that forecaster needs to use the latest evaluation results of the guidance models every year rather than rely on the well-known accuracy of models for a long time to reduce prediction error.