• The Korean Journal of Applied Statistics
• /
• v.32 no.2
• /
• pp.213-227
• /
• 2019

In this paper, we propose an end-to-end deep learning model combining Bayesian neural network with Korean speech recognition. In the past, Korean speech recognition was a complicated task due to the excessive parameters of many intermediate steps and needs for Korean expertise knowledge. Fortunately, Korean speech recognition becomes manageable with the aid of recent breakthroughs in "End-to-end" model. The end-to-end model decodes mel-frequency cepstral coefficients directly as text without any intermediate processes. Especially, Connectionist Temporal Classification loss and Attention based model are a kind of the end-to-end. In addition, we combine Bayesian neural network to implement the end-to-end model and obtain Monte Carlo estimates. Finally, we carry out our experiments on the "WorimalSam" online dictionary dataset. We obtain 4.58% Word Error Rate showing improved results compared to Google and Naver API.

• The Korean Journal of Applied Statistics
• /
• v.32 no.4
• /
• pp.517-541
• /
• 2019

According to the OECD (2015) and UN (2017), Korea was classified as an immigration country. The designation as an immigration country means that net migration will remain positive and international migration is likely to affect population growth. KOSTAT (2011) used a model with more than 15 parameters to divide sexes, immigration and emigration based on the Wilson (2010) model, which takes into account population migration factors. Five years later, we assume the average of domestic net migration rate for the last five years and foreign government policy likely quota. However, both of these results were conservative estimates of international migration and provide different results than those used by the OECD and UN to classify an immigration country. In this paper, we proposed a stochastic projection on international migration using nonparametric model (FDM by Hyndman and Ullah (2007) and Coherent FDM by Hyndman et al. (2013)) that uses a functional data model for the international migration data of Korea from 2000-2017, noting the international migration such as immigration, emigration and net migration is non-linear and not linear. According to the result, immigration rate will be 1.098(male), 1.026(female) in 2018 and 1.228(male), 1.152(female) in 2025 per 1000 population, and the emigration rate will be 0.907(male), 0.879(female) in 2018 and 0.987(male), 0.959(female) in 2025 per 1000 population. Thus the net migration is expected to increase to 0.191(male), 0.148(female) in 2018 and 0.241(male), 0.192(female) in 2025 per 1000 population.

• The Korean Journal of Applied Statistics
• /
• v.32 no.4
• /
• pp.543-559
• /
• 2019

Data often include missing values due to various reasons. If the missing data mechanism is not MCAR, analysis based on fully observed cases may an estimation cause bias and decrease the precision of the estimate since partially observed cases are excluded. Especially when data include many variables, missing values cause more serious problems. Many imputation techniques are suggested to overcome this difficulty. However, imputation methods using parametric models may not fit well with real data which do not satisfy model assumptions. In this study, we review imputation methods using nonlinear models such as kernel, resampling, and spline methods which are robust on model assumptions. In addition, we suggest utilizing imputation classes to improve imputation accuracy or adding random errors to correctly estimate the variance of the estimates in nonlinear imputation models. Performances of imputation methods using nonlinear models are compared under various simulated data settings. Simulation results indicate that the performances of imputation methods are different as data settings change. However, imputation based on the kernel regression or the penalized spline performs better in most situations. Utilizing imputation classes or adding random errors improves the performance of imputation methods using nonlinear models.

• Journal of the Korea Convergence Society
• /
• v.10 no.10
• /
• pp.153-160
• /
• 2019

In this study, The purpose of this study is to analyze how the rating scale and review contents attributes of social network-based services and products affect consumer purchasing patterns. information provided by screening the main factors. These analyzes are closely and quickly integrated between individuals and businesses, and enable to analyze the transaction that the impact of changing consumers on consumption and purchasing through the usefulness and a priori estimates of reviews and ratings at this time when networks and smart technologies are involved in a wide range of consumer activities. For this study, hierarchical analysis (AHP) and delphi (Delphi) methods applied to classify the high end variables into usefulness, technicality and value, Each subvariable was grouped into three factors and analyzed for importance through evaluation weights. As a result, we could analyze the importance of durability, usefulness, technological innovation, and cost and quality of value. Therefore, this study is expected to provide supplementary and additional useful information to consumers and companies participating in economic activities in various ways by simultaneously analyzing the review score and the reliability of posting information provided by verifying the main factors.

• Korean Journal of Veterinary Service
• /
• v.42 no.2
• /
• pp.93-112
• /
• 2019

Calf diarrhea is a common disease in young claves and is still a major cause of productivity and economic loss in livestock farms. Fecal samples from Korean native cattle (n=100) with diarrhea from 64 farms in Gwangju area, Korea from september 2017 to December 2018 were examined for shedding of important protozoan parasitic, viral and bacterial pathogens using culture, rapid test kit and PCR methods. Of 57 (89.1%) of the 64 Korean native cattle farms examined had samples infected with at least one of the investigated pathogens. Among 100 fecal samples, 88 samples were positive for at least one the twelve pathogens and 51 samples were simultaneously positive for two or more pathogens by culture and PCR assay. Bovine group A rotavirus (BRV) was the most common pathogen, found in 43/100 (43.0%) samples on 32/64 (50.0%) farms. Subsequently, kobuvirus (30.0%), pathogenic E. coli (29.0%), bovine parvovirus (17.0%), Giardia spp. (13.0%), Eimeria spp. (10.0%), Clostridium perfringens type A (8.0%), bovine torovirus (8.0%), bovine viral diarrhea virus (6.0%), bovine coronavirus (5.0%), bovine norovirus (2.0%) and Cryptosporidium spp. (2.0%) were detected. Nebovirus, kırklareli virus, bovine adenovirus, Salmonella spp. and intestinal parasites were not detected. Of the 72 calves sampled in this age group, 64 (88.9%) samples were positive for at least one enteropathogen. BRV was identified in 34/72 (47.2%) samples from 27/48 (56.3%) farms. Subsequently, pathogenic E. coli (30.6%), kobuvirus (29.2%), BPaV (22.2%), Giardia spp. (15.3%), Eimeria spp. (9.7%), BVDV (6.9%), Cl. perfringens type A (6.9%), BCoV (4.6%) and Cryptosporidium spp. (2.8%) were detected in fecal samples. A total of ninety-six strains of E. coli were isolated from one hundred fecal samples collected from Korean native cattle with diarrhea. The presence of stx1, stx2, eaeA, LT, STa, STb, ehxA, saa, F4, F5(K99), F6, F17, F18 and F41 genes in the isolates was investigated by PCR. Out of ninety-six E. coli isolates screened for specific genes, 30 strains E. coli were identified to harbor shiga toxin-producing E. coli (STEC) 7 (7.3%), enterohemorrhagic E. coli (EHEC) 8 (8.3%), enteropathogenic E. coli (EPEC) 6 (6.3%), enterotoxigenic E. coli (ETEC) 2 (2.1%) and STEC/ETEC hybrid 7 (7.3%). This study provides epidemiological estimates of the prevalence of Korean native cattle's enteropathogens in Gwangju area, Korea, which would be used for cattle farmers and veterinarians to select appropriate therapeutic method.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.19 no.12
• /
• pp.265-271
• /
• 2018

This paper presents a method for detecting damage to a structure at low cost using its impedance. The impedance technique is a typical method to detect local damage for structural health monitoring. This is a common technique for estimating damage by monitoring the electro-mechanical admittance signal of the structure. To apply this technique, an expensive impedance analyzer is generally used. On the other hand, it is necessary to develop a low-cost variant to effectively disseminate the technique. In this study, a method based on the transfer impedance using a function generator and digital multimeter, which are generally used in the laboratory instead of an impedance analyzer, was developed. That is, this technique estimates the damage by comparing the damage index using the amplitude ratio of the output voltage measured in the healthy and damaged state. A transfer impedance test was carried out on a steel specimen. By comparing the damage index, the presence of damage could be assessed reasonably. This study is a basic investigation of an impedance-based low-cost damage detection method that can be used effectively for structural health monitoring if supplemented with future research to estimate the damage location and severity.

• The Journal of the Korea Contents Association
• /
• v.21 no.10
• /
• pp.48-58
• /
• 2021

For securing technology and business competences of companies that is the engine of domestic industrial growth, government-supported policy programs for the creation of commercialization results in various forms such as 『Technology Transaction Market Vitalization』 and 『Technology Finance-based R&D Commercialization Support』 have been carried out since 2014. So far, various studies on technology valuation theories and evaluation variables have been formalized by experts from various fields, and have been utilized in the field of technology commercialization. However, Their practicality has been questioned due to the existing constraint that valuation results are assessed lower than the expectation in the evaluation sector. Even considering that the evaluation results may differ depending on factors such as the corporate situation and investment environment, it is necessary to establish a reference infrastructure to secure the objectivity and reliability of the technology valuation results. In this study, we investigate the evaluation infrastructure built by each institution and examine whether the latest artificial neural networks and deep learning technologies are applicable for performing predictive simulation of technology values based on principal variables, and predicting sales estimates and qualitative evaluation scores in order to embed onto the technology valuation system.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.2
• /
• pp.139-147
• /
• 2021

This paper introduces and implements a Sound Event Detection (SED) system based on weakly-supervised learning where only part of the data is labeled, and analyzes the effect of parameters. The SED system estimates the classes and onset/offset times of events in the acoustic signal. In order to train the model, all information on the event class and onset/offset times must be provided. Unfortunately, the onset/offset times are hard to be labeled exactly. Therefore, in the weakly-supervised task, the SED model is trained by "strongly labeled data" including the event class and activations, "weakly labeled data" including the event class, and "unlabeled data" without any label. Recently, the SED systems using the mean-teacher model are widely used for the task with several parameters. These parameters should be chosen carefully because they may affect the performance. In this paper, performance analysis was performed on parameters, such as the feature, moving average parameter, weight of the consistency cost function, ramp-up length, and maximum learning rate, using the data of DCASE 2020 Task 4. Effects and the optimal values of the parameters were discussed.

• Korean Journal of Agricultural Science
• /
• v.47 no.4
• /
• pp.845-857
• /
• 2020

The precise estimation of accelerated soil wind erosion that can cause severe economic and environmental impacts still has not been achieved to date. The objectives of this investigation were to verify the applicability of a Wind Erosion Prediction System (WEPS) that expressed the soil loss as mass per area for specific areas of interest on a daily basis for a single event in arable lands. To this end, we selected and evaluated the results published by Hagen in 2004 and the soil depth converted from the mass of soil losses obtained by using the WEPS. Hagen's results obtained from the WEPS model followed the 1 : 1 line between predicted and measured value for soil losses with only less than 2 kg·m-2 whereas the values between the measured and predicted loss did not show any correlation for the given field conditions due to the initial field surface condition although the model provided reasonable estimates of soil loss. Calculated soil depths of the soil loss by wind for both the observed and predicted ones ranged from 0.004 to 3.113 cm·10 a-1 and from 0 to 2.013 cm·10 a-1, respectively. Comparison of the soil depths between the observed and predicted ones did not show any good relationship, and there was no soil loss in the predicted one while slight soil loss was measured in the observed one. Therefore, varying the essential model inputs and factors related to wind speed and soil properties are needed to accurately estimate soil loss for a given field in arable land.

• Korean Journal of Agricultural and Forest Meteorology
• /
• v.22 no.4
• /
• pp.299-311
• /
• 2020

This paper analyzes the distribution of water vapor pressure and relative humidity in complex terrains by collecting weather observation data at 6 locations in the valley in Jungdae-ri, Ganjeon-myeon, Gurye-gun, Jeolla South Province and 14 locations in Akyang-myeon, Hadong-gun, Gyeongsang South Province, which form a single drainage basin in rural and mountainous regions. Previously estimated water vapor pressure used in the early warning system for agrometeorological hazard and actual water vapor pressure arrived at using the temperature and humidity that were measured at the highest density (1.5 m above ground) at every hour in the valley of Jungdae-ri between 19 December 2014 and 23 November 2015 and in the valley of Akyang between 15 August 2012 and 18 August 2013 were compared. The altitude-specific gradient of the observed water vapor pressure varied with different hours of the day and the difference in water vapor pressure between high and low altitudes increased in the night. The hourly variations in the water vapor pressure in the weather stations of the valley of Akyang with various topographic and ground conditions were caused by factors other than altitude. From the observed data of the study area, a coefficient that adj usts the variation in the water vapor pressure according to the specific difference in altitude and estimates it closer to the actual measured level was derived. Relative humidity was simulated as water vapor pressure estimated against the saturated water vapor pressure, thus, confirming that errors were further reduced using the derived coefficient than with the previous method that was used in the early warning system.