• Journal of Bio-Environment Control
• /
• v.29 no.2
• /
• pp.161-170
• /
• 2020

There is systematic spatial variations in environmental properties due to sensitive reaction to external conditions at plastic greenhouse occupied 99.2% of domestic agricultural facilities. In order to construct 3 dimensional distribution of temperature, relative humidity, CO₂ and illuminance, measurement matrix as 3 by 3 by 5 in direction of width, height and length, respectively, dividing indoor space of greenhouse was designed and tested at experimental site. Linear regression analysis was conducted to evaluate optimal estimation method in terms with horizontal and vertical variations. Even though sole measurement point for temperature and relative humidity could be feasible to assess indoor condition, multiple measurement matrix is inevitably required to improve spatial precision at certain time domain such as period of sunrise and sunset. In case with CO₂, multiple measurement matrix could not successfully improve the spatial predictability during a whole experimental period. In case with illuminance, prediction performance was getting smaller after a time period of sunrise due to systematic interference such as indoor structure. Thus, multiple sensing methodology was proposed in direction of length at higher height than growing bed, which could compensate estimation error in spatial domain. Appropriate measurement matrix could be constructed considering the transition of stability in indoor environmental properties due to external variations. As a result, optimal measurement matrix should be carefully designed considering flexibility of construction relevant with the type of property, indoor structure, the purpose of crop and the period of growth. For an instance, partial cooling and heating system to save a consumption of energy supplement could be successfully accomplished by the deployment of multiple measurement matrix.

• The Journal of Engineering Geology
• /
• v.18 no.4
• /
• pp.447-457
• /
• 2008

Laboratory tests for single plane sliding were conducted using the model rock slope to investigate the cut slope deformability and failure mechanism due to combined effect of engineering characteristics such as angle of sliding plane, water force, joint roughness and infillings. Also the possibility of prediction of slope failure through displacement monitoring was explored. The joint roughness was prepared in forms of saw-tooth type having different roughness specifications. The infillings was maintained between upper and lower roughness plane from zero to 1.2 times of the amplitude of the surface projections. Water force was expressed as the percent filling of tension crack from dry (0%) to full (100%), and constantly increased from 0% at the rate of 0.5%/min and 1%/min upto failure. Total of 50 tests were performed at sliding angles of $30^{\circ}$ and $35^{\circ}$ based on different combinations of joint roughness, infilling thickness and water force increment conditions. For smooth sliding plane, it was found that the linear type of deformability exhibited irrespective of the infilling thickness and water force conditions. For sliding planes having roughness, stepping or exponential types of deformability were predominant under condition that the infilling thickness is lower or higher than asperity height, respectively. These arise from the fact that, once the infilling thickness exceeds asperities, strength and deformability of the sliding plane is controlled by the engineering characteristics of the infilling materials. The results obtained in this study clearly show that the water force at failure was found to increase with increasing joint roughness, and to decrease with increasing filling thickness. It seems possible to estimate failure time using the inverse velocity method for sliding plane having exponential type of deformability. However, it is necessary to estimate failure time by trial and error basis to predict failure of the slope accurately.

• Journal of Intelligence and Information Systems
• /
• v.27 no.1
• /
• pp.65-82
• /
• 2021

Stock market investors are generally split into foreign investors, institutional investors, and individual investors. Compared to individual investor groups, professional investor groups such as foreign investors have an advantage in information and financial power and, as a result, foreign investors are known to show good investment performance among market participants. The purpose of this study is to propose an investment strategy that combines investor-specific transaction information and machine learning, and to analyze the portfolio investment performance of the proposed model using actual stock price and investor-specific transaction data. The Korea Exchange offers daily information on the volume of purchase and sale of each investor to securities firms. We developed a data collection program in C# programming language using an API provided by Daishin Securities Cybosplus, and collected 151 out of 200 KOSPI stocks with daily opening price, closing price and investor-specific net purchase data from January 2, 2007 to July 31, 2017. The self-organizing map model is an artificial neural network that performs clustering by unsupervised learning and has been introduced by Teuvo Kohonen since 1984. We implement competition among intra-surface artificial neurons, and all connections are non-recursive artificial neural networks that go from bottom to top. It can also be expanded to multiple layers, although many fault layers are commonly used. Linear functions are used by active functions of artificial nerve cells, and learning rules use Instar rules as well as general competitive learning. The core of the backpropagation model is the model that performs classification by supervised learning as an artificial neural network. We grouped and transformed investor-specific transaction volume data to learn backpropagation models through the self-organizing map model of artificial neural networks. As a result of the estimation of verification data through training, the portfolios were rebalanced monthly. For performance analysis, a passive portfolio was designated and the KOSPI 200 and KOSPI index returns for proxies on market returns were also obtained. Performance analysis was conducted using the equally-weighted portfolio return, compound interest rate, annual return, Maximum Draw Down, standard deviation, and Sharpe Ratio. Buy and hold returns of the top 10 market capitalization stocks are designated as a benchmark. Buy and hold strategy is the best strategy under the efficient market hypothesis. The prediction rate of learning data using backpropagation model was significantly high at 96.61%, while the prediction rate of verification data was also relatively high in the results of the 57.1% verification data. The performance evaluation of self-organizing map grouping can be determined as a result of a backpropagation model. This is because if the grouping results of the self-organizing map model had been poor, the learning results of the backpropagation model would have been poor. In this way, the performance assessment of machine learning is judged to be better learned than previous studies. Our portfolio doubled the return on the benchmark and performed better than the market returns on the KOSPI and KOSPI 200 indexes. In contrast to the benchmark, the MDD and standard deviation for portfolio risk indicators also showed better results. The Sharpe Ratio performed higher than benchmarks and stock market indexes. Through this, we presented the direction of portfolio composition program using machine learning and investor-specific transaction information and showed that it can be used to develop programs for real stock investment. The return is the result of monthly portfolio composition and asset rebalancing to the same proportion. Better outcomes are predicted when forming a monthly portfolio if the system is enforced by rebalancing the suggested stocks continuously without selling and re-buying it. Therefore, real transactions appear to be relevant.

• Korean Journal of Agricultural and Forest Meteorology
• /
• v.23 no.4
• /
• pp.316-328
• /
• 2021

In this study, the leaf Na content prediction model for spring potato was established using 400-1000 nm hyperspectral sensor to develop the multispectral sensor for the salinity monitoring in reclaimed land. The irrigation conditions were standard, drought, and salinity (2, 4, 8 dS/m), and the irrigation amount was calculated based on the amount of evaporation. The leaves' Na contents were measured 1st and 2nd weeks after starting irrigation in the vegetative, tuber formative, and tuber growing periods, respectively. The reflectance of the leaves was converted from 5 nm to 10 nm, 25 nm, and 50 nm of FWHM (full width at half maximum) based on the 10 nm wavelength intervals. Using the variance importance in projections of partial least square regression(PLSR-VIP), ten band ratios were selected as the variables to predict salinity damage levels with Na content of spring potato leaves. The MLR(Multiple linear regression) models were estimated by removing the band ratios one by one in the order of the lowest weight among the ten band ratios. The performance of models was compared by not only R2, MAPE but also the number of band ratios, optimal FWHM to develop the compact multispectral sensor. It was an advantage to use 25 nm of FWHM to predict the amount of Na in leaves for spring potatoes during the 1st and 2nd weeks vegetative and tuber formative periods and 2 weeks tuber growing periods. The selected bandpass filters were 15 bands and mainly in red and red-edge regions such as 430/440, 490/500, 500/510, 550/560, 570/580, 590/600, 640/650, 650/660, 670/680, 680/690, 690/700, 700/710, 710/720, 720/730, 730/740 nm.

• Journal of Korea Water Resources Association
• /
• v.54 no.spc1
• /
• pp.1095-1105
• /
• 2021

The advancement of dam operation is further required due to the upcoming rainy season, typhoons, or torrential rains. Besides, physical models based on specific rules may sometimes have limitations in controlling the release discharge of dam due to inherent uncertainty and complex factors. This study aims to forecast the water level of the nearest station to the dam multi-timestep-ahead and evaluate the availability when it makes a decision for a release discharge of dam based on LSTM (Long Short-Term Memory) of deep learning. The LSTM model was trained and tested on eight data sets with a 1-hour temporal resolution, including primary data used in the dam operation and downstream water level station data about 13 years (2009~2021). The trained model forecasted the water level time series divided by the six lead times: 1, 3, 6, 9, 12, 18-hours, and compared and analyzed with the observed data. As a result, the prediction results of the 1-hour ahead exhibited the best performance for all cases with an average accuracy of MAE of 0.01m, RMSE of 0.015 m, and NSE of 0.99, respectively. In addition, as the lead time increases, the predictive performance of the model tends to decrease slightly. The model may similarly estimate and reliably predicts the temporal pattern of the observed water level. Thus, it is judged that the LSTM model could produce predictive data by extracting the characteristics of complex hydrological non-linear data and can be used to determine the amount of release discharge from the dam when simulating the operation of the dam.

• Journal of Radiation Protection and Research
• /
• v.41 no.1
• /
• pp.41-47
• /
• 2016

Background: For the accurate dose assessment in radiation therapy, energy spectrum of the photon beam generated from the linac head is essential. The aim of this study is to develop the technique to accurately unfolding the energy spectrum with the transmission analysis method. Materials and Methods: Clinical linear accelerator and Monet Carlo method was employed to evaluate the transmission signals according to the thickness of the observer material, and then the response function of the ion chamber response was determined with the mono energy beam. Finally the energy spectrum was unfolded with HEPROW program. Elekta Synergy Flatform and Geant4 tool kits was used in this study. Results and Discussion: In the comparison between calculated and measured transmission signals using aluminum alloy as an attenuator, root mean squared error was 0.43%. In the comparison between unfolded spectrum using HEPROW program and calculated spectrum using Geant4, the difference of peak and mean energy were 0.066 and 0.03 MeV, respectively. However, for the accurate prediction of the energy spectrum, additional experiment with various type of material and improvement of the unfolding program is required. Conclusion: In this research, it is demonstrated that unfolding spectra technique could be used in megavoltage photon beam with aluminum alloy and HEPROW program.

• Fire Science and Engineering
• /
• v.15 no.1
• /
• pp.23-33
• /
• 2001

The utility tunnels are the important facility as a mainstay of country because of the latest communication developments. However, the utilities tunnel is difficult to deal with in case of a fire accident. When a cable burns, the black smoke containing poisonous gas will be reduced. This black smoke goes into the tunnel, and makes it difficult to extinguish the fire. Therefore, when there was a fire in the utility tunnel, the central nerves of the country had been paralyzed, such as property damage, communication interruption, in addition to inconvenience for people. This paper is based on the fire occurred in the past, and reenacting the fire by making the real utilities tunnel model. The aim of this paper is the scientific analysis of the character image of the fire, and the verification of each fire protection system whether it works well after process of setting up a fire protection system in the utilities tunnel at a constant temperature. The fire experiment was equipped with the linear heat detector, the fire door, the connection water spray system and the ventilation system in the utilities tunnel. Fixed portion of an electric power supply cable was coated with a fire retardant coating, and a heating tube was covered with a fireproof. The result showed that the highest temperature was $932^{\circ}c$ and the linear heat detector was working at the constant temperature, and it pointed at the place of the fire on the receiving board, and Fixed portion of the electric power supply cable coated with the fire retardant coating did not work as the fireproof. The heating tube was covered with the fireproof about 30 minutes.

• Horticultural Science & Technology
• /
• v.28 no.5
• /
• pp.768-776
• /
• 2010

This study was conducted to predict number and fresh weight of leaves, and total leaf area of a single-stemmed rose 'Vital' based on the accumulated thermal units, and to develop a model of shoot development for the prediction of the time when the flowering shoot reaches a phenological stage in a plant factory system. The base temperature ($T_b$), optimum temperature ($T_{opt}$), and maximum temperature ($T_{max}$) were estimated by regressing the rate of shoot development against the temperature gradient. The rate of shoot development ($R$, $d^{-1}$) for the phase from cutting to bud break (CT-BB) was best described by a linear model $R_b$ ($d^{-1}$) = -0.0089 + $0.0016{\cdot}temp$. The rate of shoot development for the phase from bud break to harvest (BB-HV) was fitted to the parabolic model $R_h$ ($d^{-1}$) = $-0.0001{\cdot}temp^2$ + $0.0054{\cdot}temp$ - 0.0484. The $T_b$, $T_{opt}$, and $T_{max}$ values were 5.56, 27.0, and $42.7^{\circ}C$, respectively. The $T_b$ value was used in the thermal unit computations for the shoot development. Number of leaves, leaf area (LA), and leaf fresh weight showed sigmoidal curves regardless of the cut time. The shoot development and leaf area model was described as a sigmoidal function using thermal units. Leaf area was described as LA = 578.7 $[1+(thermal units/956.1)^{-8.54}]^{-1}$. Estimated and observed shoot length and leaf fresh weight showed a reasonably good fit with 1.060 ($R^2=0.976^{***}$) and 1.043 ($R^2=0.955^{***}$), respectively. The average thermal units required from cutting to transplant and from transplant to harvest stages were $426{\pm}42^{\circ}C{\cdot}d$ and $783{\pm}24^{\circ}C{\cdot}d$, respectively.

• The korean journal of orthodontics
• /
• v.31 no.1 s.84
• /
• pp.39-50
• /
• 2001

In proceeding with orthodontic treatment, the prediction for the shape, growth rate and growth direction of mandible plays a major role to set up the treatment plan and determine its period and prognosis. Various approaches being made so far have shown that the linear and angular measurement using lateral cephalograms are relatively accurate to estimate them. This study was purposed to find the shape of mandible more clearly by preventing the overlap of the Condyle head area which appears in lateral cephalogram, and to estimate its growth rate by comparing the growth quantity and ratio via lateral area measurement. This experimental was performed against 40 patients total, of which Class I of 14, Class II of 9 and Class III of 17 consist. Wide open lateral cephalograms of 40 patients were taken over average period of 4 Year 3 Months, then the linear and angular measurements were carried out with 11 itemized lists. Autocad Rl4 application program was utilized to draw their appearance, measure and compare their lateral area. As a result of study, conclusions were made as follows; 1. Mandibular body length (gonion-menton) tended to increase in order of CIII, CI and CII, and Mandibular body length of CIII group had a tendency to grow twice faster than that of CII group. 2. In lateral items such as Go-Me, A-Cd, B-Cp, E-F and G-H, CIII showed a significant increase on the year-average quantity and rate of the growth, and especially apparent difference was observed in CIII group rather than CII group. 3. For the 4 Year 3 Months period, the year-average growth quantity of lateral area of the mandible was $1.0cm^2$ for Class I, $0.8cm^2$ for Class II and $1.4cm^2$ for Class III, which corresponds to $11.9\%,\;11.8\%\;and\;20.3\%$ of growth ratio respectively. Thus, growth ratio almost 2 times more than other groups was observed in group CIII while growth ratio between group CI and CII has little difference. 4. Considering the results as above, it can be proposed that the difference in size of the mandible between groups is caused by the difference in the growth rate and growth quantity of the mandible, which generated in the middle of growth, rather than the difference in size of congenital Jaw-bone.

• Journal of Intelligence and Information Systems
• /
• v.27 no.3
• /
• pp.175-197
• /
• 2021

Recently, with the development of deep learning technology, research on unstructured data analysis is being actively conducted, and it is showing remarkable results in various fields such as classification, summary, and generation. Among various text analysis fields, text classification is the most widely used technology in academia and industry. Text classification includes binary class classification with one label among two classes, multi-class classification with one label among several classes, and multi-label classification with multiple labels among several classes. In particular, multi-label classification requires a different training method from binary class classification and multi-class classification because of the characteristic of having multiple labels. In addition, since the number of labels to be predicted increases as the number of labels and classes increases, there is a limitation in that performance improvement is difficult due to an increase in prediction difficulty. To overcome these limitations, (i) compressing the initially given high-dimensional label space into a low-dimensional latent label space, (ii) after performing training to predict the compressed label, (iii) restoring the predicted label to the high-dimensional original label space, research on label embedding is being actively conducted. Typical label embedding techniques include Principal Label Space Transformation (PLST), Multi-Label Classification via Boolean Matrix Decomposition (MLC-BMaD), and Bayesian Multi-Label Compressed Sensing (BML-CS). However, since these techniques consider only the linear relationship between labels or compress the labels by random transformation, it is difficult to understand the non-linear relationship between labels, so there is a limitation in that it is not possible to create a latent label space sufficiently containing the information of the original label. Recently, there have been increasing attempts to improve performance by applying deep learning technology to label embedding. Label embedding using an autoencoder, a deep learning model that is effective for data compression and restoration, is representative. However, the traditional autoencoder-based label embedding has a limitation in that a large amount of information loss occurs when compressing a high-dimensional label space having a myriad of classes into a low-dimensional latent label space. This can be found in the gradient loss problem that occurs in the backpropagation process of learning. To solve this problem, skip connection was devised, and by adding the input of the layer to the output to prevent gradient loss during backpropagation, efficient learning is possible even when the layer is deep. Skip connection is mainly used for image feature extraction in convolutional neural networks, but studies using skip connection in autoencoder or label embedding process are still lacking. Therefore, in this study, we propose an autoencoder-based label embedding methodology in which skip connections are added to each of the encoder and decoder to form a low-dimensional latent label space that reflects the information of the high-dimensional label space well. In addition, the proposed methodology was applied to actual paper keywords to derive the high-dimensional keyword label space and the low-dimensional latent label space. Using this, we conducted an experiment to predict the compressed keyword vector existing in the latent label space from the paper abstract and to evaluate the multi-label classification by restoring the predicted keyword vector back to the original label space. As a result, the accuracy, precision, recall, and F1 score used as performance indicators showed far superior performance in multi-label classification based on the proposed methodology compared to traditional multi-label classification methods. This can be seen that the low-dimensional latent label space derived through the proposed methodology well reflected the information of the high-dimensional label space, which ultimately led to the improvement of the performance of the multi-label classification itself. In addition, the utility of the proposed methodology was identified by comparing the performance of the proposed methodology according to the domain characteristics and the number of dimensions of the latent label space.