• Proceedings of the IEEK Conference
• /
• 2003.11a
• /
• pp.397-400
• /
• 2003

본 연구에서는 KM-Net(Hidden Markov Network)을 다양한 태스크에의 적용과 화자의 특성을 효과적으로 나타내기 위해 HM-Net 음성인식 시스템에 MLLR(Maximum Likelihood Linear Regression) 적응방법을 도입하였으며, HM-Net 학습 알고리즘을 개량하여 회귀클래스 생성방법을 제안한다. 제안방법은 PDT-SSS(Phonetic Decision Tree-based Successive State Splitting) 알고리즘의 문맥방향 상태분할에 의한 상태레벨 공유를 이용한 방법으로 새로운 화자로부터 문맥정보와 적응화 데이터의 발성 양에 의존하여 결정된 많은 적응 파라미터들을(평균, 분산) 자유롭게 제어할 수 있게 된다. 제안방법의 유효성을 확인하기 위해 국어공학센터(KLE) 452 음성 데이터와 항공편 예약관련 연속음성을 대상으로 인식실험을 수행한 결과, 전체적으로 음소인식의 경우 평균 34-37％, 단어인식의 경우 평균 9％, 연속음성인식의 경우 평균 7-8％의 인식성능 향상을 각각 보였다. 또한 적응화 데이터의 양에 따른 인식성능 비교에서, 제안방법을 적용한 인식 시스템이 적응 데이터의 양이 적은 경우에도 향상된 인식률을 보였으며. 잡음을 부가한 음성에 대한 적응화 실험에서도 향상된 인식성능을 보여 MLLR 적응방법의 특성을 만족하였다. 따라서 MLLR 적응방법을 도입한 HM-Net 음성인식 시스템에 제안한 회귀클래스 생성방법이 유효함을 확인한 수 있었다.

• Journal of the Korea Safety Management & Science
• /
• v.8 no.3
• /
• pp.1-9
• /
• 2006

Recently data mining techniques have been used for analysis and classification of data related to industrial accidents. The main objective of this study is to compare algorithms for data analysis of industrial accidents and this paper provides an optimal predicting model of 5 kinds of algorithms including CHAID, CART, C4.5, LR (Logistic Regression) and NN (Neural Network) with ROC chart, lift chart and response threshold. Also, this paper provides an approximation model for an optimal predicting model based on NN. The approximation model provided in this study can be utilized for easy interpretation of data analysis using NN. This study uses selected ten independent variables to group injured people according to a dependent variable in a way that reduces variation. In order to find an optimal predicting model among 5 algorithms, a retrospective analysis was performed in 67,278 subjects. The sample for this work chosen from data related to industrial accidents during three years ($2002\;{\sim}\;2004$) in korea. According to the result analysis, NN has excellent performance for data analysis and classification of industrial accidents.

• Annual Conference of KIPS
• /
• 2015.10a
• /
• pp.1361-1364
• /
• 2015

본 연구에서는 노인들이 일상에서 많이 행동하는 7가지 유형의 행동의 특징을 추출하고, 총 7가지 분류 알고리즘에 적용하여 가장 인식률이 높은 알고리즘을 도출하고자 하였다. 행동패턴은 정상보행, 절름발이, 지팡이, 느린 보행, 허리가 굽은 상태에서 보행, 스스로 휠체어 끌 때 그리고 누군가가 휠체어를 끌어줄 때 총 7가지로 구성하였다. 행동패턴의 특징은 3축 가속도 센서의 값, 평균, 표준편차, 수직 및 수평축의 데이터를 사용하였다. 분류 알고리즘은 Naive Bayes, Bayes Net, k-NN, SVM, Decision Tree, Multilayer perception, Logistic regression을 사용하였다. 연구결과 k-NN 알고리즘의 인식률이 98.7%로 다른 분류알고리즘에 비해 인식률이 높게 나타났다.

• Journal of the Korean Society of School Health
• /
• v.31 no.3
• /
• pp.135-145
• /
• 2018

Purpose: The study aimed to investigate the factors associated with long hours of smartphone usage and to identify the characteristics of the high-risk group among female middle school students in South Korea. Methods: The study analyzed the data of 13,648 female middle school students using their own smartphone extracted from the 13th Youth Health Behavior Online Survey (2017). The factors related to using smartphones for a long time was analyzed by binomial logistic regression. The characteristics of the high-risk group was defined by a decision tree analysis. Results: The average hours spent on smartphone usage was 269.54 minutes per day. The significant factors associated with the long hours of smartphone usage were grade, living with parents, perceived household economic status, perceived academic achievement, stress, sadness and hopelessness, the main purpose of smartphone usage, drinking, body mass index, breakfast, and satisfaction with sleep quality. The subjects showing low academic performance and having breakfast four times a week or less were more likely to use their smartphone for a long time. Conclusion: Based on the results of the research, we need to establish intervention strategies focusing on the factors influencing long-time usage of smartphone. Particularly, the subjects who show poor academic performance and skip breakfast frequently should be considered as the high-risk group for spending long hours on smartphone usage.

• Journal of the Korean Society of Safety
• /
• v.36 no.5
• /
• pp.52-60
• /
• 2021

As the prevention of fatal accidents is considered an essential part of social responsibilities, both government and individual have devoted efforts to mitigate the unsafe conditions and behaviors that facilitate accidents. Several studies have analyzed the factors that cause fatal accidents and compared them to those of non-fatal accidents. However, studies on mathematical and systematic analysis techniques for identifying the features of fatal accidents are rare. Recently, various industrial fields have employed machine learning algorithms. This study aimed to apply machine learning algorithms for the classification of fatal and non-fatal accidents based on the features of each accident. These features were obtained by text mining literature on accidents. The classification was performed using four machine learning algorithms, which are widely used in industrial fields, including logistic regression, decision tree, neural network, and support vector machine algorithms. The results revealed that the machine learning algorithms exhibited a high accuracy for the classification of accidents into the two categories. In addition, the importance of comparing similar cases between fatal and non-fatal accidents was discussed. This study presented a method for classifying accidents using machine learning algorithms based on the reports on previous studies on accidents.

• Structural Engineering and Mechanics
• /
• v.83 no.1
• /
• pp.79-92
• /
• 2022

In this paper, the impact of a vernacular pozzolanic kaolinite mine on concrete alkali-silica reaction and strength has been evaluated. For making the samples, kaolinite powder with various levels has been used in the quality specification test of aggregates based on the ASTM C1260 standard in order to investigate the effect of kaolinite particles on reducing the reaction of the mortar bars. The compressive strength, X-Ray Diffraction (XRD) and Scanning Electron Microscope (SEM) experiments have been performed on concrete specimens. The obtained results show that addition of kaolinite powder to concrete will cause a pozzolanic reaction and decrease the permeability of concrete samples comparing to the reference concrete specimen. Further, various machine learning methods have been used to predict ASR-induced expansion per different amounts of kaolinite. In the process of modeling methods, optimal method is considered to have the lowest mean square error (MSE) simultaneous to having the highest correlation coefficient (R). Therefore, to evaluate the efficiency of the proposed model, the results of the support vector machine (SVM) method were compared with the decision tree method, regression analysis and neural network algorithm. The results of comparison of forecasting tools showed that support vector machines have outperformed the results of other methods. Therefore, the support vector machine method can be mentioned as an effective approach to predict ASR-induced expansion.

• Advances in nano research
• /
• v.12 no.5
• /
• pp.489-499
• /
• 2022

Nanofluids have recently triggered a substantial scientific interest as cooling media. However, their stability is challenging for successful engagement in industrial applications. Different factors, including temperature, nanoparticles and base fluids characteristics, pH, ultrasonic power and frequency, agitation time, and surfactant type and concentration, determine the nanofluid stability regime. Indeed, it is often too complicated and even impossible to accurately find the conditions resulting in a stabilized nanofluid. Furthermore, there are no empirical, semi-empirical, and even intelligent scenarios for anticipating the stability of nanofluids. Therefore, this study introduces a straightforward and reliable intelligent classifier for discriminating among the stability regimes of alumina-water nanofluids based on the Zeta potential margins. In this regard, various intelligent classifiers (i.e., deep learning and multilayer perceptron neural network, decision tree, GoogleNet, and multi-output least squares support vector regression) have been designed, and their classification accuracy was compared. This comparison approved that the multilayer perceptron neural network (MLPNN) with the SoftMax activation function trained by the Bayesian regularization algorithm is the best classifier for the considered task. This intelligent classifier accurately detects the stability regimes of more than 90% of 345 different nanofluid samples. The overall classification accuracy and misclassification percent of 90.1% and 9.9% have been achieved by this model. This research is the first try toward anticipting the stability of water-alumin nanofluids from some easily measured independent variables.

• International Journal of Computer Science & Network Security
• /
• v.22 no.5
• /
• pp.330-334
• /
• 2022

In light of the decreasing crop production and shortage of food across the world, one of the crucial criteria of agriculture nowadays is selecting the right crop for the right piece of land at the right time. First problem is that How Farmers can predict the right crop for cultivation because famers have no knowledge about prediction of crop. Second problem is that which algorithm is best that provide the maximum accuracy for crop prediction. Therefore, in this research Author proposed a method that would help to select the most suitable crop(s) for a specific land based on the analysis of the affecting parameters (Temperature, Humidity, Soil Moisture) using machine learning. In this work, the author implemented Random Forest Classifier, Support Vector Machine, k-Nearest Neighbor, and Decision Tree for crop selection. The author trained these algorithms with the training dataset and later these algorithms were tested with the test dataset. The author compared the performances of all the tested methods to arrive at the best outcome. In this way best algorithm from the mention above is selected for crop prediction.

• Journal of Korean Dental Science
• /
• v.16 no.1
• /
• pp.47-62
• /
• 2023

Purpose: We aimed to investigate the objective cutoff values of unstimulated flow rates (UFR) and stimulated salivary flow rates (SFR) in patients with xerostomia and to present an optimal machine learning model with a classification and regression tree (CART) for all ages. Materials and Methods: A total of 829 patients with oral diseases were enrolled (591 females; mean age, 59.29±16.40 years; 8~95 years old), 199 patients with xerostomia and 630 patients without xerostomia. Salivary and clinical characteristics were collected and analyzed. Result: Patients with xerostomia had significantly lower levels of UFR (0.29±0.22 vs. 0.41±0.24 ml/min) and SFR (1.12±0.55 vs. 1.39±0.94 ml/min) (P<0.001), respectively, compared to those with non-xerostomia. The presence of xerostomia had a significantly negative correlation with UFR (r=-0.603, P=0.002) and SFR (r=-0.301, P=0.017). In the diagnosis of xerostomia based on the CART algorithm, the presence of stomatitis, candidiasis, halitosis, psychiatric disorder, and hyperlipidemia were significant predictors for xerostomia, and the cutoff ranges for xerostomia for UFR and SFR were 0.03~0.18 ml/min and 0.85~1.6 ml/min, respectively. Conclusion: Xerostomia was correlated with decreases in UFR and SFR, and their cutoff values varied depending on the patient's underlying oral and systemic conditions.

• Computers and Concrete
• /
• v.32 no.3
• /
• pp.233-246
• /
• 2023

Accurate prediction of concrete compressive strength can minimize the need for extensive, time-consuming, and costly mixture optimization testing and analysis. This study attempts to enhance the prediction accuracy of compressive strength using stacking ensemble machine learning (ML) with feature engineering techniques. Seven alternative ML models of increasing complexity were implemented and compared, including linear regression, SVM, decision tree, multiple layer perceptron, random forest, Xgboost and Adaboost. To further improve the prediction accuracy, a ML pipeline was proposed in which the feature engineering technique was implemented, and a two-layer stacked model was developed. The k-fold cross-validation approach was employed to optimize model parameters and train the stacked model. The stacked model showed superior performance in predicting concrete compressive strength with a correlation of determination (R²) of 0.985. Feature (i.e., variable) importance was determined to demonstrate how useful the synthetic features are in prediction and provide better interpretability of the data and the model. The methodology in this study promotes a more thorough assessment of alternative ML algorithms and rather than focusing on any single ML model type for concrete compressive strength prediction.