• Smart Structures and Systems
• /
• v.33 no.5
• /
• pp.325-332
• /
• 2024

This study proposes a novel long short-term memory (LSTM)-based approach for predicting carbonation depth, with the aim of enhancing the durability evaluation of concrete structures. Conventional carbonation depth prediction relies on statistical methodologies using carbonation influencing factors and in-situ carbonation depth data. However, applying in-situ data for predictive modeling faces challenges due to the lack of time-series data. To address this limitation, an LSTM-based carbonation depth prediction technique is proposed. First, training data are generated through random sampling from the distribution of carbonation velocity coefficients, which are calculated from in-situ carbonation depth data. Subsequently, a Bayesian theorem is applied to tailor the training data for each target bridge, which are depending on surrounding environmental conditions. Ultimately, the LSTM model predicts the time-dependent carbonation depth data for the target bridge. To examine the feasibility of this technique, a carbonation depth dataset from 3,960 in-situ bridges was used for training, and untrained time-series data from the Miho River bridge in the Republic of Korea were used for experimental validation. The results of the experimental validation demonstrate a significant reduction in prediction error from 8.19% to 1.75% compared with the conventional statistical method. Furthermore, the LSTM prediction result can be enhanced by sequentially updating the LSTM model using actual time-series measurement data.

• Journal of Digital Contents Society
• /
• v.18 no.3
• /
• pp.525-534
• /
• 2017

The open set recognition method should be used for the cases that the classes of test data are not known completely in the training phase. So it is required to include two processes of classification and the validation test. This kind of research is very necessary for commercialization of face recognition modules, but few domestic researches results about it have been published. In this paper, we propose an open set face recognition method that includes two sequential validation phases. In the first phase, with dummy classes we perform classification based on sparse representation. Here, when the test data is classified into a dummy class, we conclude that the data is invalid. If the data is classified into one of the regular training classes, for second validation test we extract four features and apply them for the proposed decision function. In experiments, we proposed a simulation method for open set recognition and showed that the proposed validation test outperform SCI of the well-known validation method

• Korean Journal of Remote Sensing
• /
• v.38 no.4
• /
• pp.395-409
• /
• 2022

The purpose of this study is to evaluate the classification performance and applicability when land cover datasets constructed for AI training are cross validation to other areas. For study areas, Gyeongsang-do and Jeolla-do in South Korea were selected as cross validation areas, and training datasets were obtained from AI-Hub. The obtained datasets were applied to the U-Net algorithm, a semantic segmentation algorithm, for each region, and the accuracy was evaluated by applying them to the same and other test areas. There was a difference of about 13-15% in overall classification accuracy between the same and other areas. For rice field, fields and buildings, higher accuracy was shown in the Jeolla-do test areas. For roads, higher accuracy was shown in the Gyeongsang-do test areas. In terms of the difference in accuracy by weight, the result of applying the weights of Gyeongsang-do showed high accuracy for forests, while that of applying the weights of Jeolla-do showed high accuracy for dry fields. The result of land cover classification, it was found that there is a difference in classification performance of existing datasets depending on area. When constructing land cover map for AI training, it is expected that higher quality datasets can be constructed by reflecting the characteristics of various areas. This study is highly scalable from two perspectives. First, it is to apply satellite images to AI study and to the field of land cover. Second, it is expanded based on satellite images and it is possible to use a large scale area and difficult to access.

• Journal of The Korean Society of Agricultural Engineers
• /
• v.66 no.1
• /
• pp.1-13
• /
• 2024

This study assessed the efficacy of improving the accuracy of reservoir water level prediction models by employing automated machine learning models and efficient cross-validation methods for time-series data. Considering the inherent complexity and non-linearity of time-series data related to reservoir water levels, we proposed an optimized approach for model selection and training. The performance of twelve models was evaluated for the Obong Reservoir in Gangneung, Gangwon Province, using the TPOT (Tree-based Pipeline Optimization Tool) and four cross-validation methods, which led to the determination of the optimal pipeline model. The pipeline model consisting of Extra Tree, Stacking Ridge Regression, and Simple Ridge Regression showed outstanding predictive performance for both training and test data, with an R² (Coefficient of determination) and NSE (Nash-Sutcliffe Efficiency) exceeding 0.93. On the other hand, for predictions of water levels 12 hours later, the pipeline model selected through time-series split cross-validation accurately captured the change pattern of time-series water level data during the test period, with an NSE exceeding 0.99. The methodology proposed in this study is expected to greatly contribute to the efficient generation of reservoir water level predictions in regions with high rainfall variability.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.05a
• /
• pp.516-518
• /
• 2022

In this paper, we introduce a dependence of number of nodes of hidden-layer in fall detection system using Long Short-Term Memory that can detect falls. Its training is carried out using the parameter theta(θ), which indicates the angle formed by the x, y, and z-axis data for the direction of gravity using a 3-axis acceleration sensor. In its learning, validation is performed and divided into training data and test data in a ratio of 8:2, and training is performed by changing the number of nodes in the hidden layer to increase efficiency. When the number of nodes is 128, the best accuracy is shown with Accuracy = 99.82%, Specificity = 99.58%, and Sensitivity = 100%.

• Journal of the Korean Data and Information Science Society
• /
• v.15 no.2
• /
• pp.499-506
• /
• 2004

The Unscented Kalman Filter is known to outperform the Extended Kalman Filter for the nonlinear state estimation with a significance advantage that it does not require the computation of Jacobian but EKF has a competitive advantage to the UKF on the performance time. We compare both algorithms on training the artificial neural network. The validation data set is used to estimate parameters which are supposed to result in better fitting for the test data set. Experimental results are presented which indicate the performance of both algorithms.

• Journal of the Korean Data and Information Science Society
• /
• v.27 no.2
• /
• pp.549-557
• /
• 2016

In this paper we propose multiclassification method for large data sets by ensembling least squares support vector machines (LS-SVM) with principal components instead of raw input vector. We use the revised one-vs-all method for multiclassification, which is one of voting scheme based on combining several binary classifications. The revised one-vs-all method is performed by using the hat matrix of LS-SVM ensemble, which is obtained by ensembling LS-SVMs trained using each random sample from the whole large training data. The leave-one-out cross validation (CV) function is used for the optimal values of hyper-parameters which affect the performance of multiclass LS-SVM ensemble. We present the generalized cross validation function to reduce computational burden of leave-one-out CV functions. Experimental results from real data sets are then obtained to illustrate the performance of the proposed multiclass LS-SVM ensemble.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2019.05a
• /
• pp.156-156
• /
• 2019

In this study, a gated recurrent unit (GRU) network is constructed based on a deep neural network (DNN) with the aim of restoring the missing daily flow data in river basins. Lai Chau hydrological station is located upstream of the Da river basin (Vietnam) is selected as the target station for this study. Input data of the model are data on observed daily flow for 24 years from 1961 to 1984 (before Hoa Binh dam was built) at 5 hydrological stations, in which 4 gauge stations in the basin downstream and restoring - target station (Lai Chau). The total available data is divided into sections for different purposes. The data set of 23 years (1961-1983) was employed for training and validation purposes, with corresponding rates of 80% for training and 20% for validation respectively. Another data set of one year (1984) was used for the testing purpose to objectively verify the performance and accuracy of the model. Though only a modest amount of input data is required and furthermore the Lai Chau hydrological station is located upstream of the Da River, the calculated results based on the suggested model are in satisfactory agreement with observed data, the Nash - Sutcliffe efficiency (NSE) is higher than 95%. The finding of this study illustrated the outstanding performance of the GRU network model in recovering the missing flow data at Lai Chau station. As a result, DNN models, as well as GRU network models, have great potential for application within the field of hydrology and hydraulics.

• Journal of Korea Multimedia Society
• /
• v.23 no.11
• /
• pp.1349-1360
• /
• 2020

In recent years, Convolutional Neural Networks (CNNs) have been successfully implemented in different tasks of computer vision. Since CNN models are the representatives of supervised learning algorithms, they demand large amount of data in order to train the classifiers. Thus, obtaining data with correct labels is imperative to attain the state-of-the-art performance of the CNN models. However, labelling datasets is quite tedious and expensive process, therefore real-life datasets often exhibit incorrect labels. Although the issue of poorly labelled datasets has been studied before, we have noticed that the methods are very complex and hard to reproduce. Therefore, in this research work, we propose Deep CleanNet - a considerably simple system that achieves competitive results when compared to the existing methods. We use K-means clustering algorithm for selecting data with correct labels and train the new dataset using a deep CNN model. The technique achieves competitive results in both training and validation stages. We conducted experiments using MNIST database of handwritten digits with 50% corrupted labels and achieved up to 10 and 20% increase in training and validation sets accuracy scores, respectively.

• Communications for Statistical Applications and Methods
• /
• v.13 no.1
• /
• pp.151-165
• /
• 2006

It is important to estimate the true misclassification rate of a given classifier when an independent set of test data is not available. Cross-validation and bootstrap are two possible approaches in this case. In related literature bootstrap estimators of the true misclassification rate were asserted to have better performance for small samples than cross-validation estimators. We compare the two estimators empirically when the classification rule is so adaptive to training data that its apparent misclassification rate is close to zero. We confirm that bootstrap estimators have better performance for small samples because of small variance, and we have found a new fact that their bias tends to be significant even for moderate to large samples, in which case cross-validation estimators have better performance with less computation.