• Korean Journal of Radiology
• /
• v.24 no.2
• /
• pp.155-165
• /
• 2023

Objective: Little is known about the effects of using different expert-determined reference standards when evaluating the performance of deep learning-based automatic detection (DLAD) models and their added value to radiologists. We assessed the concordance of expert-determined standards with a clinical gold standard (herein, pathological confirmation) and the effects of different expert-determined reference standards on the estimates of radiologists' diagnostic performance to detect malignant pulmonary nodules on chest radiographs with and without the assistance of a DLAD model. Materials and Methods: This study included chest radiographs from 50 patients with pathologically proven lung cancer and 50 controls. Five expert-determined standards were constructed using the interpretations of 10 experts: individual judgment by the most experienced expert, majority vote, consensus judgments of two and three experts, and a latent class analysis (LCA) model. In separate reader tests, additional 10 radiologists independently interpreted the radiographs and then assisted with the DLAD model. Their diagnostic performance was estimated using the clinical gold standard and various expert-determined standards as the reference standard, and the results were compared using the t test with Bonferroni correction. Results: The LCA model (sensitivity, 72.6%; specificity, 100%) was most similar to the clinical gold standard. When expert-determined standards were used, the sensitivities of radiologists and DLAD model alone were overestimated, and their specificities were underestimated (all p-values < 0.05). DLAD assistance diminished the overestimation of sensitivity but exaggerated the underestimation of specificity (all p-values < 0.001). The DLAD model improved sensitivity and specificity to a greater extent when using the clinical gold standard than when using the expert-determined standards (all p-values < 0.001), except for sensitivity with the LCA model (p = 0.094). Conclusion: The LCA model was most similar to the clinical gold standard for malignant pulmonary nodule detection on chest radiographs. Expert-determined standards caused bias in measuring the diagnostic performance of the artificial intelligence model.

• Journal of the Korea Society of Computer and Information
• /
• v.29 no.5
• /
• pp.93-100
• /
• 2024

With the advent of the Fourth Industrial Revolution, cutting-edge technologies such as artificial intelligence, big data, the Internet of Things, and cloud computing are driving innovation across industries. These technologies are generating massive amounts of data that many companies are leveraging. However, there is a notable reluctance among users to share sensitive information due to the privacy risks associated with collecting personal data. This is particularly evident in the healthcare sector, where the collection of sensitive information such as patients' medical conditions poses significant challenges, with privacy concerns hindering data collection and analysis. This research presents a novel technique for collecting and analyzing medical data that not only preserves privacy, but also effectively extracts statistical information. This method goes beyond basic data collection by incorporating a strategy to efficiently mine statistical data while maintaining privacy. Performance evaluations using real-world data have shown that the propose technique outperforms existing methods in extracting meaningful statistical insights.

• Atmosphere
• /
• v.34 no.2
• /
• pp.177-185
• /
• 2024

As increasing global interest in renewable energy due to the ongoing climate crisis, there is a growing need for efficient technologies to manage such resources. This study focuses on the predictive skill of daily solar power generation using weather observation and forecast data. Meteorological data from the Korea Meteorological Administration and solar power generation data from the Korea Power Exchange were utilized for the period from January 2017 to May 2023, considering both inland (Daejeon) and coastal (Busan) regions. Temperature, wind speed, relative humidity, and precipitation were selected as relevant meteorological variables for solar power prediction. All data was preprocessed by removing their systematic components to use only their residuals and the residual of solar data were further processed with weighted adjustments for homoscedasticity. Four models, MLR (Multiple Linear Regression), RF (Random Forest), DNN (Deep Neural Network), and RNN (Recurrent Neural Network), were employed for solar power prediction and their performances were evaluated based on predicted values utilizing observed meteorological data (used as a reference), 1-day-ahead forecast data (referred to as fore1), and 2-day-ahead forecast data (fore2). DNN-based prediction model exhibits superior performance in both regions, with RNN performing the least effectively. However, MLR and RF demonstrate competitive performance comparable to DNN. The disparities in the performance of the four different models are less pronounced than anticipated, underscoring the pivotal role of fitting models using residuals. This emphasizes that the utilized preprocessing approach, specifically leveraging residuals, is poised to play a crucial role in the future of solar power generation forecasting.

• Informatization Policy
• /
• v.31 no.2
• /
• pp.39-64
• /
• 2024

This study assesses the national AI competitiveness of 38 OECD countries with focus on AI human capital, AI infrastructure, and AI innovation capacity. Utilizing the fuzzy-set ideal type analysis method, these countries were categorized into eight distinct types based on their national AI competitiveness levels, leading to the derivation of pertinent implications. The analysis identified a category termed "AI Leading Country" consisting of North American, Western European, and Nordic countries, along with several Asian nations including South Korea. Remarkably, the United States demonstrated dominant global national AI competitiveness, achieving the highest fuzzy scores across all three evaluative factors. South Korea was classified as an "AI Leading Country" primarily due to its superior AI infrastructure, but its performance in AI human capital and AI innovation capacity was found to be moderate relative to other analyzed nations; thus highlighting the necessity of sustained focus on the accumulation of AI human capital and bolstering of AI innovation capacity.

• Journal of Advanced Navigation Technology
• /
• v.28 no.3
• /
• pp.356-361
• /
• 2024

This paper deals with the design and fabrication of a coplanar waveguide (CPW)-fed super-wideband semicircular-disk-shaped dipole antenna operating in a frequency band of 2.4 GHz or higher. To feed the antenna, a CPW feed line was appended to the center of the lower arm of the semicircular-disk-shaped dipole antenna. For miniaturization, square patches were added to the ends of the two arms of the semicircular-disk-shaped dipole, whereas the slot width of the CPW feed line at the center of the dipole antenna was increased to improve impedance matching in the 5.4-6.3 GHz band. The simulated frequency band of the proposed antenna for a voltage standing wave ratio (VSWR) less than 2 was 2.369-30 GHz(170.7%), whereas the fabricated antenna was maintained VSWR less than 2 in the frequency range of 2.378-20 GHz when measured using a network analyzer operating up to 20 GHz so it can be applied as a super-wideband antenna for next-generation mobile communications.

• Steel and Composite Structures
• /
• v.51 no.6
• /
• pp.679-695
• /
• 2024

This paper presents a novel technique that combines machine learning (ML) with moth-flame optimization (MFO) methods to predict the axial compressive strength (ACS) of concrete filled double skin steel tubes (CFDST) columns. The proposed model is trained and tested with a dataset containing 125 tests of the CFDST column subjected to compressive loading. Five ML models, including extreme gradient boosting (XGBoost), gradient tree boosting (GBT), categorical gradient boosting (CAT), support vector machines (SVM), and decision tree (DT) algorithms, are utilized in this work. The MFO algorithm is applied to find optimal hyperparameters of these ML models and to determine the most effective model in predicting the ACS of CFDST columns. Predictive results given by some performance metrics reveal that the MFO-CAT model provides superior accuracy compared to other considered models. The accuracy of the MFO-CAT model is validated by comparing its predictive results with existing design codes and formulae. Moreover, the significance and contribution of each feature in the dataset are examined by employing the SHapley Additive exPlanations (SHAP) method. A comprehensive uncertainty quantification on probabilistic characteristics of the ACS of CFDST columns is conducted for the first time to examine the models' responses to variations of input variables in the stochastic environments. Finally, a web-based application is developed to predict ACS of the CFDST column, enabling rapid practical utilization without requesting any programing or machine learning expertise.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.22 no.6
• /
• pp.182-195
• /
• 2023

Recently, research on traffic signal control using artificial intelligence algorithms has been receiving attention, and many traffic signal control models are being studied. However, most studies either focused on independent intersections or are theoretical studies that calculate signal cycle length according to changes in traffic volume. Therefore, this study was conducted on a signalized intersection - roundabout in Gajwa-ro. The Particle Swarm Optimization - Bacterial Foraging Optimization (PSO-BFO) algorithm was proposed, which is developed from the GA and PSO algorithms for minimizing congestion at two intersections. As a result, optimum cycle length was determined to be 158 seconds. The Verkehr In Stadten - SIMulationsmodell (VISSIM) results showed that there was 3.4% increased capacity, 8.2% reduced delay and 8.3% reduced number of stops at the Gajwa-ro signalized intersection. Additionally, at the roundabout, a 9.2% increase in capacity, a 7.1% reduction in delay, and a 27.2% decrease in the number of stops was observed.

• The Journal of the Convergence on Culture Technology
• /
• v.10 no.2
• /
• pp.465-470
• /
• 2024

The advancement of AI-based image generation technology has resulted in the creation of various images, emphasizing the need for technology capable of accurately discerning them. The amount of generated image data is limited, and to achieve high performance with a limited dataset, this study proposes a model for discriminating generated images using transfer learning. Applying pre-trained models from the ImageNet dataset directly to the CIFAKE input dataset, we reduce training time cost followed by adding three hidden layers and one output layer to fine-tune the model. The modeling results revealed an improvement in the performance of the model when adjusting the final layer. Using transfer learning and then adjusting layers close to the output layer, small image data-related accuracy issues can be reduced and generated images can be classified.

• Analyses & Alternatives
• /
• v.8 no.1
• /
• pp.59-85
• /
• 2024

This study examines the impact of the development of the artificial intelligence (AI) industry on the economic growth of South Korea. The study uses variables such as the revenue and patent applications of AI-related companies, as well as industry-specific total factor productivity and GDP, to estimate the effects. The results suggest that the growth of the AI industry has a positive effect on the economic growth with a lag of about one year. Specifically, the effect of government AI revenue on GDP growth appears to be greater than that of private companies or consumer-focused AI revenue. This indicates that government policies aimed at promoting the diffusion of the AI industry have had significant effects. The study notes that the period covered by the AI industry survey data is relatively short, and there is a lack of detailed data for the manufacturing sector. I suggest that further improvements and accumulation of data could lead to more robust results.

• Geomechanics and Engineering
• /
• v.37 no.1
• /
• pp.49-64
• /
• 2024

Rocks undergoing repeated loading and unloading over an extended period, such as due to earthquakes, human excavation, and blasting, may result in the gradual accumulation of stress and deformation within the rock mass, eventually reaching an unstable state. In this study, a CNN-CCM is proposed to address the mechanical behavior. The structure and hyperparameters of CNN-CCM include Conv2D layers × 5; Max pooling2D layers × 4; Dense layers × 4; learning rate=0.001; Epoch=50; Batch size=64; Dropout=0.5. Training and validation data for deep learning include 71 rock samples and 122,152 data points. The AI Rock Constitutive Model learned by CNN-CCM can predict strain values(ε₁) using Mass (M), Axial stress (σ₁), Density (ρ), Cyclic number (N), Confining pressure (σ₃), and Young's modulus (E). Five evaluation indicators R², MAPE, RMSE, MSE, and MAE yield respective values of 0.929, 16.44%, 0.954, 0.913, and 0.542, illustrating good predictive performance and generalization ability of model. Finally, interpreting the AI Rock Constitutive Model using the SHAP explaining method reveals that feature importance follows the order N > M > σ₁ > E > ρ > σ₃.Positive SHAP values indicate positive effects on predicting strain ε₁ for N, M, σ₁, and σ₃, while negative SHAP values have negative effects. For E, a positive value has a negative effect on predicting strain ε₁, consistent with the influence patterns of conventional physical rock constitutive equations. The present study offers a novel approach to the investigation of the mechanical constitutive model of rocks under cyclic loading and unloading conditions.