• Journal of Information Processing Systems
• /
• v.16 no.4
• /
• pp.809-819
• /
• 2020

Gender prediction accuracy increases as convolutional neural network (CNN) architecture evolves. This paper compares voting and ensemble schemes to utilize the already trained five CNN models to further improve gender prediction accuracy. The majority voting usually requires odd-numbered models while the proposed softmax-based voting can utilize any number of models to improve accuracy. The ensemble of CNN models combined with one more fully-connected layer requires further tuning or training of the models combined. With experiments, it is observed that the voting or ensemble of CNN models leads to further improvement of gender prediction accuracy and that especially softmax-based voters always show better gender prediction accuracy than majority voters. Also, compared with softmax-based voters, ensemble models show a slightly better or similar accuracy with added training of the combined CNN models. Softmax-based voting can be a fast and efficient way to get better accuracy without further training since the selection of the top accuracy models among available CNN pre-trained models usually leads to similar accuracy to that of the corresponding ensemble models.

• KEPCO Journal on Electric Power and Energy
• /
• v.7 no.1
• /
• pp.115-123
• /
• 2021

Through machine learning-based load prediction, it is possible to prevent excessive power generation or unnecessary economic investment by estimating the appropriate amount of facility investment in consideration of the load that will increase in the future or providing basic data for policy establishment to distribute the maximum load. However, in order to secure the reliability of the developed load prediction model in the field, the performance comparison verification between the distribution line load prediction models must be preceded, but a comparative performance verification system between the distribution line load prediction models has not yet been established. As a result, it is not possible to accurately determine the performance excellence of the load prediction model because it is not possible to easily determine the likelihood between the load prediction models. In this paper, we developed a reliability verification system for load prediction models including a method of comparing and verifying the performance reliability between machine learning-based load prediction models that were not previously considered, verification process, and verification result visualization methods. Through the developed load prediction model reliability verification system, the objectivity of the load prediction model performance verification can be improved, and the field application utilization of an excellent load prediction model can be increased.

• Journal of information and communication convergence engineering
• /
• v.21 no.4
• /
• pp.346-350
• /
• 2023

Trajectory prediction is an essential element for driving autonomous vehicles, and various trajectory prediction models have emerged with the development of deep learning technology. Convolutional neural network (CNN) is the most commonly used neural network architecture for extracting the features of visual images, and the latest models exhibit high performances. This study was conducted to identify an efficient CNN backbone model among the components of deep learning models for trajectory prediction. We changed the existing CNN backbone network of multiple-trajectory prediction models used as feature extractors to various state-of-the-art CNN models. The experiment was conducted using nuScenes, which is a dataset used for the development of autonomous vehicles. The results of each model were compared using frequently used evaluation metrics for trajectory prediction. Analyzing the impact of the backbone can improve the performance of the trajectory prediction task. Investigating the influence of the backbone on multiple deep learning models can be a future challenge.

• Structural Engineering and Mechanics
• /
• v.74 no.4
• /
• pp.503-514
• /
• 2020

This study proposes prediction models for the shear strength of steel fiber reinforced concrete (SFRC) and ultra-high-performance fiber reinforced concrete (UHPC) beams using a Bayesian parameter estimation approach and a collected experimental database. Previous researchers had already proposed shear strength prediction models for SFRC and UHPC beams, but their performances were limited in terms of their prediction accuracies and the applicability to UHPC beams. Therefore, this study adopted a statistical approach based on a collected database to develop prediction models. In the database, 89 and 37 experimental data for SFRC and UHPC beams without stirrups were collected, respectively, and the proposed equations were developed using the Bayesian parameter estimation approach. The proposed models have a simplified form with important parameters, and in comparison to the existing prediction models, provide unbiased high prediction accuracy.

• Atmosphere
• /
• v.21 no.2
• /
• pp.209-220
• /
• 2011

A new consensus algorithm for the prediction of tropical cyclone track has been developed. Conventional consensus is a simple average of a few fixed models that showed the good performance in track prediction for the past few years. Meanwhile, the consensus in this study is a weighted average of a few models that may change for every individual forecast time. The models are selected as follows. The first step is to find the analogous past tropical cyclone tracks to the current track. The next step is to evaluate the model performances for those past tracks. Finally, we take the weighted average of the selected models. More weight is given to the higher performance model. This new algorithm has been named as DYTRAP (DYnamic data-base Typhoon tRAck Prediction) in the sense that the data base is used to find the analogous past tracks and the effective models for every individual track prediction case. DYTRAP has been applied to all 2009 tropical cyclone track prediction. The results outperforms those of all models as well as all the official forecasts of the typhoon centers. In order to prove the real usefulness of DYTRAP, it is necessary to apply the DYTRAP system to the real time prediction because the forecast in typhoon centers usually uses 6-hour or 12-hour-old model guidances.

• Journal of Information Technology Services
• /
• v.11 no.1
• /
• pp.211-221
• /
• 2012

Many metric-based classification models have been proposed to predict fault-proneness of software module. This paper presents two prediction models using Bayesian classifier which is one of the most popular modern classification algorithms. Bayesian model based on Bayesian probability theory can be a promising technique for software quality prediction. This is due to the ability to represent uncertainty using probabilities and the ability to partly incorporate expert's knowledge into training data. The two models, Na$\ddot{i}$veBayes(NB) and Bayesian Belief Network(BBN), are constructed and dimensionality reduction of training data and test data are performed before model evaluation. Prediction accuracy of the model is evaluated using two prediction error measures, Type I error and Type II error, and compared with well-known prediction models, backpropagation neural network model and support vector machine model. The results show that the prediction performance of BBN model is slightly better than that of NB. For the data set with ambiguity, although the BBN model's prediction accuracy is not as good as the compared models, it achieves better performance than the compared models for the data set without ambiguity.

• Journal of Korean Society of Transportation
• /
• v.17 no.1
• /
• pp.141-157
• /
• 1999

This paper first reviews current practices of traveler information providing and provides some perspectives regarding the possible near term milestones in traveler information providing. Then, reviews of four types of prediction models: 1) dynamic traffic assignment (DTA) model; 2) statistical model; 3) simulation model; and 4) heuristic model are described in the sense that various prediction models are needed to support providing predictive traveler information in the near future. Next, the functional requirements and capabilities of the four types of prediction models are discussed and summarized along with some advantages and disadvantages of these models with reference to short-term travel time prediction. Furthermore, a comprehensive prediction procedure, which combines the four types of prediction models, is presented, together with the data requirements for each type of prediction model.

• Nuclear Engineering and Technology
• /
• v.54 no.8
• /
• pp.3027-3033
• /
• 2022

Background: In this study, various types of deep-learning models for predicting in vitro radiosensitivity from gene-expression profiling were compared. Methods: The clonogenic surviving fractions at 2 Gy from previous publications and microarray gene-expression data from the National Cancer Institute-60 cell lines were used to measure the radiosensitivity. Seven different prediction models including three distinct multi-layered perceptrons (MLP), four different convolutional neural networks (CNN) were compared. Folded cross-validation was applied to train and evaluate model performance. The criteria for correct prediction were absolute error < 0.02 or relative error < 10%. The models were compared in terms of prediction accuracy, training time per epoch, training fluctuations, and required calculation resources. Results: The strength of MLP-based models was their fast initial convergence and short training time per epoch. They represented significantly different prediction accuracy depending on the model configuration. The CNN-based models showed relatively high prediction accuracy, low training fluctuations, and a relatively small increase in the memory requirement as the model deepens. Conclusion: Our findings suggest that a CNN-based model with moderate depth would be appropriate when the prediction accuracy is important, and a shallow MLP-based model can be recommended when either the training resources or time are limited.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.45 no.2
• /
• pp.20-29
• /
• 2022

This study intends to present a traffic node-based and link-based accident prediction models using XGBoost which is very excellent in performance among machine learning models, and to develop those models with sustainability and scalability. Also, we intend to present those models which predict the number of annual traffic accidents based on road types, weather conditions, and traffic information using XGBoost. To this end, data sets were constructed by collecting and preprocessing traffic accident information, road information, weather information, and traffic information. The SHAP method was used to identify the variables affecting the number of traffic accidents. The five main variables of the traffic node-based accident prediction model were snow cover, precipitation, the number of entering lanes and connected links, and slow speed. Otherwise, those of the traffic link-based accident prediction model were snow cover, precipitation, the number of lanes, road length, and slow speed. As the evaluation results of those models, the RMSE values of those models were each 0.2035 and 0.2107. In this study, only data from Sejong City were used to our models, but ours can be applied to all regions where traffic nodes and links are constructed. Therefore, our prediction models can be extended to a wider range.

• Korean Journal of Remote Sensing
• /
• v.17 no.3
• /
• pp.199-210
• /
• 2001

The purpose of this study is to compare the landslide prediction models quantitatively using prediction rate curve. A case study from the Jangheung area was used to illustrate the methodologies. The landslide locations were detected from remote sensing data and field survey, and geospatial information related to landslide occurrences were built as a spatial database in GIS. As prediction models, joint conditional probability model and certainty factor model were applied. For cross-validation approach, landslide locations were partitioned into two groups randomly. One group was used to construct prediction models, and the other group was used to validate prediction results. From the cross-validation analysis, it is possible to compare two models to each other in this study area. It is expected that these approaches will be used effectively to compare other prediction models and to analyze the causal factors in prediction models.