• Journal of KIISE:Software and Applications
• /
• v.34 no.12
• /
• pp.1045-1055
• /
• 2007

Latent topic models are statistical models which automatically captures salient patterns or correlation among features underlying a data collection in a probabilistic way. They are gaining an increased popularity as an effective tool in the application of automatic semantic feature extraction from text corpus, multimedia data analysis including image data, and bioinformatics. Among the important issues for the effectiveness in the application of latent topic models to the massive data set is the efficient learning of the model. The paper proposes an accelerated learning technique for PLSA model, one of the popular latent topic models, by an incremental EM algorithm instead of conventional EM algorithm. The incremental EM algorithm can be characterized by the employment of a series of partial E-steps that are performed on the corresponding subsets of the entire data collection, unlike in the conventional EM algorithm where one batch E-step is done for the whole data set. By the replacement of a single batch E-M step with a series of partial E-steps and M-steps, the inference result for the previous data subset can be directly reflected to the next inference process, which can enhance the learning speed for the entire data set. The algorithm is advantageous also in that it is guaranteed to converge to a local maximum solution and can be easily implemented just with slight modification of the existing algorithm based on the conventional EM. We present the basic application of the incremental EM algorithm to the learning of PLSA and empirically evaluate the acceleration performance with several possible data partitioning methods for the practical application. The experimental results on a real-world news data set show that the proposed approach can accomplish a meaningful enhancement of the convergence rate in the learning of latent topic model. Additionally, we present an interesting result which supports a possible synergistic effect of the combination of incremental EM algorithm with parallel computing.

• Proceedings of the Korean Geotechical Society Conference
• /
• 2010.03a
• /
• pp.188-197
• /
• 2010

This paper presents experimental results of a series of 1-g shaking table model tests performed on end-bearing single piles and pile groups to investigate the effect of particle size on the dynamic behavior of soil-pile systems. Two soil-pile models consisting of a single-pile and a $4{\times}2$-pile group were tested twice; first using Jumoonjin sand, and second using Australian Fine sand, which has a smaller particle size. In the case of single-pile models, the lateral displacement was almost within 1% of pile diameter which corresponds to the elastic range of the pile. The back-calculated p-y curves show that the subgrade reaction of the Jumoonjin-sand-model ground was larger than that of the Australian Fine-sand-model ground at the same displacement. This phenomenon means that the stress-strain behavior of Jumoonjin sand was initially stiffer than that of Australian Fine sand. This difference was also confirmed by resonant column tests and compression triaxial tests. And the single pile p-y backbone curves of the Australian fine sand were constructed and compared with those of the Jumoonjin sand. As a result, the stiffness of the p-y backbone curves of Jumunjin sand was larger than those of Australian fine sand. Therefore, using the same p-y curves regardless of particle size can lead to inaccurate results when evaluating dynamic behavior of soil-pile system. In the case of the group-pile models, the lateral displacement was much larger than the elastic range of pile movement at the same test conditions in the single-pile models. The back-calculated p-y curves in the case of group pile models were very similar in both sands because the stiffness difference between the Jumoonjin-sand-model ground and the Australian Fine-sand-model ground was not significantly large at a large strain level, where both sands showed non-linear behavior. According to a series of single pile and group pile test results, the evaluation group pile effect using the p-multiplier can lead to inaccurate results on dynamic behavior of soil-pile system.

• Journal of Korean Society of Coastal and Ocean Engineers
• /
• v.7 no.3
• /
• pp.265-276
• /
• 1995

In this Paper, the Performance of the convolution method has been investigated as an effort to develop a simple system of predicting wind-driven surface current on a real time basis. In this approach wind stress is assumed to be spatially uniform and the effect of atmospheric pressure is neglected. The discrete convolution weights are determined in advance at each point using a linear three-dimensional Galerkin model with linear shape functions(Galerkin-FEM model). Four directions of wind stress(e.g. NE, SW, NW, SE) with unit magnitude are imposed in the model calculation for the construction of data base for convolution weights. Given the time history of wind stress, it is then possible to predict with-driven currents promptly using the convolution product of finite length. An unsteady wind stress of arbitrary form can be approximated by a series of wind pulses with magnitude of 6 hour averaged value. A total of 12 pulses are involved in the convolution product To examine the accuracy of the convolution method a series of numerical experiments has been carried out in the idealized basin representing the scale of the Yellow Sea and the East China Sea. The wind stress imposed varies sinusoidally in time. It was found that the predicted surface currents and elevation fields were in good agreement with the results computed by the direct integration of the Galerkin model. A model with grid 1/8$^{\circ}$ in latitude, l/6$^{\circ}$ in longitude was established which covers the entire region of the Yellow Sea and the East China Sea. The numerical prediction in terms of the convolution product has been carried out with particular attention on the formation of upwind flow in the middle of the Yellow Sea by northerly wind.

• Journal of Distribution Science
• /
• v.11 no.10
• /
• pp.63-72
• /
• 2013

Purpose - This study analyzes the relationship and dynamic interactions between stock price index, interest rate, price index, and housing price indices using Korean monthly data from 2000 to 2013, based on a VAR model. This study also examines Granger causal relationships among these variables in order to determine whether the time series of one is useful in forecasting another, or to infer certain types of causal dependency between stochastic variables. Research design, data, and methodology - We used Korean monthly data for all variables from 2000: M1 to 2013: M3. First, we checked the correlations among different variables. Second, we conducted the Augmented Dickey-Fuller (ADF) test and the co-integration test using the VAR model. Third, we employed Granger Causality tests to quantify the causal effect from time series observations. Fourth, we used the impulse response function and variance decomposition based on the VAR model to examine the dynamic relationships among the variables. Results - First, stock price Granger affects interest rate and all housing price indices. Price index Granger, in turn, affects the stock price and six metropolitan housing price indices. However, none of the Granger variables affect the price index. Therefore, it is the stock markets (and not the housing market) that affects the housing prices. Second, the impulse response tests show that maximum influence on stock price is its own, and though it is influenced a little by interest rate, price index affects it negatively. One standard deviation (S.D.) shock to stock price increases the housing price by 0.08 units after two months, whereas an impulse shock to the interest rate negatively impacts the housing price. Third, the variance decomposition results report that the shock to the stock price accounts for 96% of the variation in the stock price, and the shock to the price index accounts for 2.8% after two periods. In contrast, the shock to the interest rate accounts for 80% of the variation in the interest rate after ten periods; the shock to the stock price accounts for 19% of the variation; however, shock to the price index does not affect the interest rate. The housing price index in 10 periods is explained up to 96.7% by itself, 2.62% by stock price, 0.68% by price index, and 0.04% by interest rate. Therefore, the housing market is explained most by its own variation, whereas the interest rate has little impact on housing price. Conclusions - The results of the study elucidate the relationship and dynamic interactions among stock price index, interest rate, price index, and housing price indices using VAR model. This study could help form the basis for more appropriate economic policies in the future. As the housing market is very important in Korean economy, any changes in house price affect the other markets, thereby resulting in a shock to the entire economy. Therefore, the analysis on the dynamic relationships between the housing market and economic variables will help with the decision making regarding the housing market policy.

• Journal of the Korea Society for Simulation
• /
• v.24 no.2
• /
• pp.31-39
• /
• 2015

Unemployment is related to social issues as well as personal economics activity so various policies have been made to reduce the unemployment rate in many countries. Because of delay inherent in the survey mechanism to collect unemployment data, it takes lots of time to acquire survey unemployment data. To develop proper policies for reducing unemployment rate at the right time, it is quite critical to obtain faster and more accurate information concerning about unemployment level. To remedy this problem, recently an advanced analytics utilizing internet queries is suggested. To examine the potential of Web query information, this research investigates the usefulness of internet activity data to predict Korean unemployment rate. One of selected web-query data(unemployment claim) has a quite strong correlation with unemployment rate. This research employes a time series approach of the ARIMA model that utilizes the information of keyword queries provided by the Naver(Korean representative portal site) trend together with unemployment rate data provisioned from Statistics Korea. With respect to model selection guidelines of mean squared error and prediction error, the model with utilizing the web query information shows better results than the model without such information. This suggests that there is a strong potential for the used method, which needs to be further explored.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.7 no.2
• /
• pp.79-88
• /
• 1987

This study is intended to propose a system reliability analysis model for R.C. bridge superstructures based on the systems reliability theory. Approximately assuming that the ultimate capacity of the superstructures is reached, when two adjacent girders fail subsequently, a practical system reliability model is proposed, which is based on a point estimate for Level II parallel-series system modelling. The sensitivity analysis of system reliabilities for the variation of the coefficients of correlations between the failure modes is performed by applying the proposed model for R.C. T beam bridges. It is observed that the point estimate method for the proposed model corresponds to the average value of the Ditlevsen's bound, and the system reliability index, ${\beta}_s$, varies quite sensitively according to the variation of the cofficients of correlations. Systems reliabilities of a few existing T beam bridges are analyzed by applying the proposed practical system reliability method of this study, and, in addition, the preferable direction of the development of the reliability-based code calibration using the system target reliability index concept are suggested.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.12 no.3
• /
• pp.242-250
• /
• 2019

In this paper, we propose a method to predict the failure of industrial robot using Seq2Seq (Sequence to Sequence) model, which is a model for transforming time series data among Artificial Neural Network models. The proposed method uses the data of the joint current and angular value, which can be measured by the robot itself, without additional sensor for fault diagnosis. After preprocessing the measured data for the model to learn, the Seq2Seq model was trained to convert the current to angle. Abnormal degree for fault diagnosis uses RMSE (Root Mean Squared Error) during unit time between predicted angle and actual angle. The performance evaluation of the proposed method was performed using the test data measured under different conditions of normal and defective condition of the robot. When the Abnormal degree exceed the threshold, it was classified as a fault, and the accuracy of the fault diagnosis was 96.67% from the experiment. The proposed method has the merit that it can perform fault prediction without additional sensor, and it has been confirmed from the experiment that high diagnostic performance and efficiency are available without requiring deep expert knowledge of the robot.

• Journal of Korean Tunnelling and Underground Space Association
• /
• v.21 no.2
• /
• pp.227-242
• /
• 2019

This paper presents a method to predict ground types ahead of a tunnel face utilizing operational data of the earth pressure-balanced (EPB) shield tunnel boring machine (TBM) when running through soil ground. The time series analysis model which was applicable to predict the mixed ground composed of soils and rocks was modified to be applicable to soil tunnels. Using the modified model, the feasibility on the choice of the soil conditioning materials dependent upon soil types was studied. To do this, a self-organizing map (SOM) clustering was performed. Firstly, it was confirmed that the ground types should be classified based on the percentage of 35% passing through the #200 sieve. Then, the possibility of predicting the ground types by employing the modified model, in which the TBM operational data were analyzed, was studied. The efficacy of the modified model is demonstrated by its 98% accuracy in predicting ground types ten rings ahead of the tunnel face. Especially, the average prediction accuracy was approximately 93% in areas where ground type variations occur.

• Ocean Science Journal
• /
• v.43 no.4
• /
• pp.195-208
• /
• 2008

This paper evaluates whether a thermodynamic ocean-carbon model can be used to predict the monthly mean global fields of the surface-water partial pressure of $CO_2$ ($pCO_{2SEA}$) from sea surface salinity (SSS), temperature (SST), and/or nitrate ($NO_3$) concentration using previously published regional total inorganic carbon ($C_T$) and total alkalinity ($A_T$) algorithms. The obtained $pCO_{2SEA}$ values and their amplitudes of seasonal variability are in good agreement with multi-year observations undertaken at the sites of the Bermuda Atlantic Timeseries Study (BATS) ($31^{\circ}50'N$, $60^{\circ}10'W$) and the Hawaiian Ocean Time-series (HOT) ($22^{\circ}45'N$, $158^{\circ}00'W$). By contrast, the empirical models predicted $C_T$ less accurately at the Kyodo western North Pacific Ocean Time-series (KNOT) site ($44^{\circ}N$, $155^{\circ}E$) than at the BATS and HOT sites, resulting in greater uncertainties in $pCO_{2SEA}$ predictions. Our analysis indicates that the previously published empirical $C_T$ and $A_T$ models provide reasonable predictions of seasonal variations in surface-water $pCO_{2SEA}$ within the (sub) tropical oceans based on changes in SSS and SST; however, in high-latitude oceans where ocean biology affects $C_T$ to a significant degree, improved $C_T$ algorithms are required to capture the full biological effect on $C_T$ with greater accuracy and in turn improve the accuracy of predictions of $pCO_{2SEA}$.

• Economic and Environmental Geology
• /
• v.19 no.4
• /
• pp.297-304
• /
• 1986

Earthquake disaster is dependent upon both site intensity and strength of structures. The higher the strength, structures become more safe, which in turn increases the construction cost. Therefore, it is necessary to decide an optimum design intensity in which the safety is balanced with the cost. Such an optimum design intensity for major man-made structures in Korea is determined in the present study from a simulation model as follows. 1) Hypothetical earthquake time series are generated from the probability distribution to represent appropriately the seismicity of Korea. 2) The strength of structures constructed with a certain design intensity is assumed to exponentially decrease with the elapsed time. The construction cost is also expressed as a function of design intensity. 3) Comparing the seismic intensity generated from the earthquake time series with the strength of structures, the safety of structures is examined. Then the time until the structure is damaged by an earthquake is obtained within the designed life time. 4) The above simulation is iterated several hundred times and hence the mean life time of structures having a certain design intensity is obtained. 5) After all, the optimum design intensity to minimize the annual mean loss, the ratio of construction cost to mean life time, is estimated. The major conclusions obtained from the above simulation model are as follows. 1) Depending upon the designed life time ($T_p$), the optimum design intensities are appeared to be 0. 05-0. 10g for $T_p=50yr$ and 0. 08-0.13g for $T_p=100yr$. 2) According to the sensitivity analysis, the optimum design intensity increases with the rapid strength decrease of structure and decreases with the increase of initial construction cost.