• Journal of Intelligence and Information Systems
• /
• v.18 no.2
• /
• pp.29-45
• /
• 2012

Bond rating is regarded as an important event for measuring financial risk of companies and for determining the investment returns of investors. As a result, it has been a popular research topic for researchers to predict companies' credit ratings by applying statistical and machine learning techniques. The statistical techniques, including multiple regression, multiple discriminant analysis (MDA), logistic models (LOGIT), and probit analysis, have been traditionally used in bond rating. However, one major drawback is that it should be based on strict assumptions. Such strict assumptions include linearity, normality, independence among predictor variables and pre-existing functional forms relating the criterion variablesand the predictor variables. Those strict assumptions of traditional statistics have limited their application to the real world. Machine learning techniques also used in bond rating prediction models include decision trees (DT), neural networks (NN), and Support Vector Machine (SVM). Especially, SVM is recognized as a new and promising classification and regression analysis method. SVM learns a separating hyperplane that can maximize the margin between two categories. SVM is simple enough to be analyzed mathematical, and leads to high performance in practical applications. SVM implements the structuralrisk minimization principle and searches to minimize an upper bound of the generalization error. In addition, the solution of SVM may be a global optimum and thus, overfitting is unlikely to occur with SVM. In addition, SVM does not require too many data sample for training since it builds prediction models by only using some representative sample near the boundaries called support vectors. A number of experimental researches have indicated that SVM has been successfully applied in a variety of pattern recognition fields. However, there are three major drawbacks that can be potential causes for degrading SVM's performance. First, SVM is originally proposed for solving binary-class classification problems. Methods for combining SVMs for multi-class classification such as One-Against-One, One-Against-All have been proposed, but they do not improve the performance in multi-class classification problem as much as SVM for binary-class classification. Second, approximation algorithms (e.g. decomposition methods, sequential minimal optimization algorithm) could be used for effective multi-class computation to reduce computation time, but it could deteriorate classification performance. Third, the difficulty in multi-class prediction problems is in data imbalance problem that can occur when the number of instances in one class greatly outnumbers the number of instances in the other class. Such data sets often cause a default classifier to be built due to skewed boundary and thus the reduction in the classification accuracy of such a classifier. SVM ensemble learning is one of machine learning methods to cope with the above drawbacks. Ensemble learning is a method for improving the performance of classification and prediction algorithms. AdaBoost is one of the widely used ensemble learning techniques. It constructs a composite classifier by sequentially training classifiers while increasing weight on the misclassified observations through iterations. The observations that are incorrectly predicted by previous classifiers are chosen more often than examples that are correctly predicted. Thus Boosting attempts to produce new classifiers that are better able to predict examples for which the current ensemble's performance is poor. In this way, it can reinforce the training of the misclassified observations of the minority class. This paper proposes a multiclass Geometric Mean-based Boosting (MGM-Boost) to resolve multiclass prediction problem. Since MGM-Boost introduces the notion of geometric mean into AdaBoost, it can perform learning process considering the geometric mean-based accuracy and errors of multiclass. This study applies MGM-Boost to the real-world bond rating case for Korean companies to examine the feasibility of MGM-Boost. 10-fold cross validations for threetimes with different random seeds are performed in order to ensure that the comparison among three different classifiers does not happen by chance. For each of 10-fold cross validation, the entire data set is first partitioned into tenequal-sized sets, and then each set is in turn used as the test set while the classifier trains on the other nine sets. That is, cross-validated folds have been tested independently of each algorithm. Through these steps, we have obtained the results for classifiers on each of the 30 experiments. In the comparison of arithmetic mean-based prediction accuracy between individual classifiers, MGM-Boost (52.95%) shows higher prediction accuracy than both AdaBoost (51.69%) and SVM (49.47%). MGM-Boost (28.12%) also shows the higher prediction accuracy than AdaBoost (24.65%) and SVM (15.42%)in terms of geometric mean-based prediction accuracy. T-test is used to examine whether the performance of each classifiers for 30 folds is significantly different. The results indicate that performance of MGM-Boost is significantly different from AdaBoost and SVM classifiers at 1% level. These results mean that MGM-Boost can provide robust and stable solutions to multi-classproblems such as bond rating.

• KIPS Transactions on Computer and Communication Systems
• /
• v.4 no.6
• /
• pp.185-196
• /
• 2015

Energy aware server clusters aim to reduce power consumption at maximum while keeping QoS(Quality of Service) compared to energy non-aware server clusters. They adjust the power mode of each server in a fixed or variable time interval to let only the minimum number of servers needed to handle current user requests ON. Previous studies on energy aware server cluster put efforts to reduce power consumption further or to keep QoS, but they do not consider energy efficiency well. In this paper, we propose an energy efficient cluster management based on autonomous learning for energy aware server clusters. Using parameters optimized through autonomous learning, our method adjusts server power mode to achieve maximum performance with respect to power consumption. Our method repeats the following procedure for adjusting the power modes of servers. Firstly, according to the current load and traffic pattern, it classifies current workload pattern type in a predetermined way. Secondly, it searches learning table to check whether learning has been performed for the classified workload pattern type in the past. If yes, it uses the already-stored parameters. Otherwise, it performs learning for the classified workload pattern type to find the best parameters in terms of energy efficiency and stores the optimized parameters. Thirdly, it adjusts server power mode with the parameters. We implemented the proposed method and performed experiments with a cluster of 16 servers using three different kinds of load patterns. Experimental results show that the proposed method is better than the existing methods in terms of energy efficiency: the numbers of good response per unit power consumed in the proposed method are 99.8%, 107.5% and 141.8% of those in the existing static method, 102.0%, 107.0% and 106.8% of those in the existing prediction method for banking load pattern, real load pattern, and virtual load pattern, respectively.

• Journal of Intelligence and Information Systems
• /
• v.23 no.2
• /
• pp.39-70
• /
• 2017

Data-drive analytics techniques have been recently applied to public surveys. Instead of simply gathering survey results or expert opinions to research the preference for a recently launched product, enterprises need a way to collect and analyze various types of online data and then accurately figure out customer preferences. In the main concept of existing data-based survey methods, the sentiment lexicon for a particular domain is first constructed by domain experts who usually judge the positive, neutral, or negative meanings of the frequently used words from the collected text documents. In order to research the preference for a particular product, the existing approach collects (1) review posts, which are related to the product, from several product review web sites; (2) extracts sentences (or phrases) in the collection after the pre-processing step such as stemming and removal of stop words is performed; (3) classifies the polarity (either positive or negative sense) of each sentence (or phrase) based on the sentiment lexicon; and (4) estimates the positive and negative ratios of the product by dividing the total numbers of the positive and negative sentences (or phrases) by the total number of the sentences (or phrases) in the collection. Furthermore, the existing approach automatically finds important sentences (or phrases) including the positive and negative meaning to/against the product. As a motivated example, given a product like Sonata made by Hyundai Motors, customers often want to see the summary note including what positive points are in the 'car design' aspect as well as what negative points are in thesame aspect. They also want to gain more useful information regarding other aspects such as 'car quality', 'car performance', and 'car service.' Such an information will enable customers to make good choice when they attempt to purchase brand-new vehicles. In addition, automobile makers will be able to figure out the preference and positive/negative points for new models on market. In the near future, the weak points of the models will be improved by the sentiment analysis. For this, the existing approach computes the sentiment score of each sentence (or phrase) and then selects top-k sentences (or phrases) with the highest positive and negative scores. However, the existing approach has several shortcomings and is limited to apply to real applications. The main disadvantages of the existing approach is as follows: (1) The main aspects (e.g., car design, quality, performance, and service) to a product (e.g., Hyundai Sonata) are not considered. Through the sentiment analysis without considering aspects, as a result, the summary note including the positive and negative ratios of the product and top-k sentences (or phrases) with the highest sentiment scores in the entire corpus is just reported to customers and car makers. This approach is not enough and main aspects of the target product need to be considered in the sentiment analysis. (2) In general, since the same word has different meanings across different domains, the sentiment lexicon which is proper to each domain needs to be constructed. The efficient way to construct the sentiment lexicon per domain is required because the sentiment lexicon construction is labor intensive and time consuming. To address the above problems, in this article, we propose a novel product reputation mining algorithm that (1) extracts topics hidden in review documents written by customers; (2) mines main aspects based on the extracted topics; (3) measures the positive and negative ratios of the product using the aspects; and (4) presents the digest in which a few important sentences with the positive and negative meanings are listed in each aspect. Unlike the existing approach, using hidden topics makes experts construct the sentimental lexicon easily and quickly. Furthermore, reinforcing topic semantics, we can improve the accuracy of the product reputation mining algorithms more largely than that of the existing approach. In the experiments, we collected large review documents to the domestic vehicles such as K5, SM5, and Avante; measured the positive and negative ratios of the three cars; showed top-k positive and negative summaries per aspect; and conducted statistical analysis. Our experimental results clearly show the effectiveness of the proposed method, compared with the existing method.

• KOREAN JOURNAL OF CROP SCIENCE
• /
• v.51 no.5
• /
• pp.386-395
• /
• 2006

The core questions for determining nitrogen topdress rate (Npi) at panicle initiation stage (PIS) are 'how much nitrogen accumulation during the reproductive stage (PNup) is required for the target rice yield or protein content depending on the growth and nitrogen nutrition status at PIS?' and 'how can we diagnose the growth and nitrogen nutrition status easily at real time basis?'. To address these questions, two years experiments from 2001 to 2002 were done under various rates of basal, tillering, and panicle nitrogen fertilizer by employing a rice cultivar, Hwaseongbyeo. The response of grain yield and milled-rice protein content was quantified in relation to RVIgreen (green ratio vegetation index) and SPAD reading measured around PIS as indirect estimators for growth and nitrogen nutrition status, the regression models were formulated to predict PNup based on the growth and nitrogen nutrition status and Npi at PIS. Grain yield showed quadratic response to PNup, RVIgreen around PIS, and SPAD reading around PIS. The regression models to predict grain yield had a high determination coefficient of above 0.95. PNup for the maximum grain yield was estimated to be 9 to 13.5 kgN/10a within the range of RVIgreen around PIS of this experiment. decreasing with increasing RVIgreen and also to be 10 to 11 kgN/10a regardless of SPAD readings around PIS. At these PNup's the protein content of milled rice was estimated to rise above 9% that might degrade eating quality seriously Milled-rice protein content showed curve-linear increase with the increase of PNup, RVIgreen around PIS, and SPAD reading around PIS. The regression models to predict protein content had a high determination coefficient of above 0.91. PNup to control the milled-rice protein content below 7% was estimated as 6 to 8 kgN/10a within the range of RVIgreen and SPAD reading of this experiment, showing much lower values than those for the maximum grain yield. The recovery of the Npi applied at PIS ranged from 53 to 83%, increasing with the increased growth amount while decreasing with the increasing Npi. The natural nitrogen supply from PIS to harvest ranged from 2.5 to 4 kg/10a, showing quadratic relationship with the shoot dry weight or shoot nitrogen content at PIS. The regression models to estimate PNup was formulated using Npi and anyone of RVIgreen, shoot dry weight, and shoot nitrogen content at PIS as predictor variables. These models showed good fitness with determination coefficients of 0.86 to 0.95 The prescription method based on the above models predicting grain yield, protein content and PNup and its constraints were discussed.

• Journal of Korean Society of Environmental Engineers
• /
• v.32 no.1
• /
• pp.53-60
• /
• 2010

In the present study, experiments have been performed to investigate the possibility of removing Cochlodinium polykrikoides using the dredged sediment from a coastal fishery and then to derive the optimal conditions; the amount and particle size of dredged sediment besprinkled into water, the thermal treatment, the types and amounts of additives, and the depth profile of Cochlodinium polykrikoides. Results showed that the optimal amount of dredged sediment besprinkled into water was 6~10 g/L, and the removal efficiency of Cochlodinium polykrikoides after the reaction time for 60 min was 73~93%. Note that, in the real sea water, it is necessary to besprinkle 6~10 $kg/m^3$ of dry dredged sediment on a unit area (1 $m^2$). With decreasing particle size, Cochlodinium polykrikoides could be more efficiently removed. The removal efficiency was 93% with the dredged sediment smaller than 100 ${\mu}m$, whereas it was 51% with that of 100 ${\mu}m$ ${\mu}m$. Since most of dredged sediment (over 90%) was smaller than 100 ${\mu}m$, high efficiency could be obtained by besprinkling only the dredged sediment without pre-treatment. CaO was found to be an effective additive in promoting the removal efficiency (up to 99%). The optimal amount of additive was 5~10%, however, it was necessary to use as small amount of an additive as possible in order to avoid the sharp increase in pH. The removal efficiency increased with increasing depth profile of Cochlodinium polykrikoides. The removal efficiency was 83% at 5 cm depth, whereas it was 93% at 50 cm depth. In the sea water, red tide occurred within 3 m depth, and furthermore most Cochlodinium polykrikoides existed within 1 m depth. It was, therefore, expected that higher removal efficiency of Cochlodinium polykrikoides could be obtained when the dredged sediment was besprinkled into the sea water. The removal efficiency of Cochlodinium polykrikoides was up to 93% when the dredged sediment (<100 ${\mu}m$) was besprinkled into water at the ratio of 10 g/L. This result was comparable to that obtained with loess (90~97%). All the results in the present study indicated that the dredged sediment from a coastal fishery could be successfully used as a substitute of loess for removing the red tide alga.

• Hwankyungkyoyuk
• /
• v.7 no.1
• /
• pp.46-87
• /
• 1994

Our country has been consuming a huge amount of energy in the course of industrialization and its demand is expected to increase enormously in the future. However, the deposits of energy resources are so limited that the settlement of energy problem comes up the essential subject. To solve the energy problem, it is requested that new resources to gain energy stably should be developed and also energy should be economized and used effectively. The effective use of energy and an the wisdom of economy in energy are requested to everybody and these things should be habitualized from very young age through education. Nevertheless, almost every school in our country hasn’t been concerned about energy education. Even though they have a concern, they are very short of the energy education materials and the quality of the materials is not so good. Therefore it is very meaningful to the settlement of energy problem of the country to make the students who will lead our country to make the students who will lead our country in the future realize the seriousness of energy problem and to provide them the necessary knowledge and methods to solve this problem so that they practice those things in everyday life. Having these necessities, this research, supported by The Korea Energy Management Corporation(KEMCO), was performed for 8 months from April 17, 1994 to December 17, 1994. Many peoples participated in this study such as 30 staffs of researchers and authors, 5 staffs of photographers and illustrators, and 3 VCR program producers developing an energy education material set for middle school students that includes a printed material for student, a diskette for computer simulation, a teacher's guidebook, VCR material and its guidebook. The following main development direction was established : First, the material for student should be consisted of units that let students know the seriousness of energy problem. Second, the focus should be put on the necessary method and practice to economize energy actually in real life based on the basic knowledge learned in elementary school. Third, material for student should be consisted of modules to be student activity-oriented teaching-learning rather than lecture-oriented one. The activity, to maximize student's interests, should be presented in various forms such as experiments, investigation, play, data interpretation, computer simulation, visits, expression and appreciation, etc. To develop the energy education materials for middle school students, a research plan was made first. After literature review about domestic and foreign energy education materials, several research trips home and abroad, and discussion meetings, the basic theory of energy education such as the principle, objective, contents, teaching-learning method, and evaluation method was established. Material for student was developed through the following procedures : The activities in the existing energy education materials were analysed and were divided into four categories related to energy using places of home, school, community, and country, and which were again divided into three categories related to time of past, present, and future, Considering these division, nine modules which are structure units of material for student were chosen, Each module comprises 2-4 activities. Totally 31 activities were designed in this way. The syllabi were made out for each activity and writing was asked for to experts related to each activity after several discussions and revision. To complement the draft, another several discussions and revision were also made on it and then pictures and illustrations were asked for. All these procedures complete the material for student, titled ; Energy Inquiry of Middle School Students', which totals 129 pages and is all in color. As the manuscript of material for student was fixed, writing for teacher's guidebook was asked for to the same writers. The draft of teacher's guidebook was also complemented through the several concentrated works and discussions. Teacher's guidebook focused on the teaching-learning principle and methods of energy education and on the concrete instruction cases for effective instruction of material for student. It is organized with two parts : the one is 'general outline' which introduces theoretical contents and the other is 'details' which are practically helpful to teaching-learning. It is totally 131 pages including both 'general outline' and 'details'. The VCR material and its guidebook consist of contents that cultivate the good attitude trying to economize energy and raise student's interests with a purpose of strong motivation to recognize the necessity of economy and practice it. After establishing development direction of VCR material through discussion meetings and research trips, its script was made by relevant experts. Then the script was also reviewed two times. The drafted VCR material made by a video material developing expert was examined and modified by previews twice. After completion of VCR material, the VCR guidebook was made. All these procedures led to the development of VCR material which runs 20 minutes in VHS type. The VCR guidebook shows a production purpose of the program, structure of contents, evaluation methods, and contents of the program in detail to give help to instructors when they use this VCR material, When these energy education materials are used, it is desirable that the VCR material should be presented first to induce student's motive, and then material for student is introduced Since the material for student is composed of activity-oriented modules and each module is independent one another in general, and each activity is, too. the necessary module or activity can be chosen and utilized in any order according to school or class conditions. This energy education materials will contribute to the development of student's ability to solve energy problem in everyday life and teacher's ability to teach the fundamental knowledge and method in solving energy problem.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.17 no.2
• /
• pp.87-95
• /
• 2016

This study examined the characteristics of the hourly demand of city gas in Korea and proposed multiple regression models to obtain precise estimates of the hourly demand of city gas. Forecasting the hourly demand of city gas with accuracy is essential in terms of safety and cost. If underestimated, the pipeline pressure needs to be increased sharply to meet the demand, when safety matters. In the opposite case, unnecessary inventory and operation costs are incurred. Data analysis showed that the hourly demand of city gas has a very high autocorrelation and that the 24-hour demand pattern of a day follows the previous 24-hour demand pattern of the same day. That is, there is a weekly cycle pattern. In addition, some conditions that temperature affects the hourly demand level were found. That is, the absolute value of the correlation coefficient between the hourly demand and temperature is about 0.853 on average, while the absolute value of the correlation coefficient on a specific day improves to 0.861 at worst and 0.965 at best. Based on this analysis, this paper proposes a multiple regression model incorporating the hourly demand ahead of 24 hours and the hourly demand ahead of 168 hours, and another multiple regression model with temperature as an additional independent variable. To show the performance of the proposed models, computational experiments were carried out using real data of the domestic city gas demand from 2009 to 2013. The test results showed that the first regression model exhibits a forecasting accuracy of MAPE (Mean Absolute Percentage Error) around 4.5% over the past five years from 2009 to 2013, while the second regression model exhibits 5.13% of MAPE for the same period.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.18 no.12
• /
• pp.126-134
• /
• 2017

There have been numerous studies on extracting the R-peak from electrocardiogram (ECG) signals. However, most of the detection methods are complicated to implement in a real-time portable electrocardiograph device and have the disadvantage of requiring a large amount of calculations. R-peak detection requires pre-processing and post-processing related to baseline drift and the removal of noise from the commercial power supply for ECG data. An adaptive filter technique is widely used for R-peak detection, but the R-peak value cannot be detected when the input is lower than a threshold value. Moreover, there is a problem in detecting the P-peak and T-peak values due to the derivation of an erroneous threshold value as a result of noise. We propose a robust R-peak detection algorithm with low complexity and simple computation to solve these problems. The proposed scheme removes the baseline drift in ECG signals using an adaptive filter to solve the problems involved in threshold extraction. We also propose a technique to extract the appropriate threshold value automatically using the minimum and maximum values of the filtered ECG signal. To detect the R-peak from the ECG signal, we propose a threshold neighborhood search technique. Through experiments, we confirmed the improvement of the R-peak detection accuracy of the proposed method and achieved a detection speed that is suitable for a mobile system by reducing the amount of calculation. The experimental results show that the heart rate detection accuracy and sensitivity were very high (about 100%).

• Journal of the Korean Society of Fisheries and Ocean Technology
• /
• v.35 no.2
• /
• pp.136-146
• /
• 1999

The telemetry system for the oxygen, pH, turbidity and the distribution ecology of fishes was constructed by the authors in order to product and manage effectively in shallow sea culture and setnets fisheries, and then the experiments for the telemetry system carried out at the culturing fishing ground in coast of Sanyang-Myon, Kyoungsangnam-Do and the set net fishing ground located Nungpo bay in Kojedo province respectively from October, 1997 to June 1998.As those results, the techniques suggested in the telemetry system for which find out the relationship between the physical and chemical environment in the sea and the distribution ecology of fishes gave full display its function, and its system could be operated as real time system. This research can also provide base-line data to develope a hybrid system unifying the marine environment information and the fisheries resources information in order to manage effectively coastal fishing ground.

• Korean Journal of Soil Science and Fertilizer
• /
• v.42 no.2
• /
• pp.79-87
• /
• 2009

Pot experiments using sand culture were conducted in 2004 under greenhouse conditions to evaluate the effect of nitrogen deficiency on red pepper biomass. Nitrogen stress was imposed by implementing 6 levels (40% to 140%) of N in Hoagland's nutrient solution for red pepper. Canopy reflectance measurements were made with hand held spectral sensors including $GreenSeeker^{TM}$, $Crop\;Circle^{TM}$, and $Field\;Scout^{TM}$ Chlorophyll meter, and a spectroradiometer as well as Minolta SPAD-502 chlorophyll meter. Canopy reflectance and dry weight of red pepper were measured at five growth stages, the 30th, 40th, 50th, 80th and 120th day after planting(DAT). Dry weight of red pepper affected by nitrogen stress showed large differences between maximum and minimum values at the 120th DAT ranged from 48.2 to $196.6g\;plant^{-1}$, respectively. Several reflectance indices obtained from $GreenSeeker^{TM}$, $Crop\;Circle^{TM}$ and Spectroradiometer including chlorophyll readings were compared for evaluation of red pepper biomass. The reflectance indices such as rNDVI, aNDVI and gNDVI by the $Crop\;Circle^{TM}$ sensor showed the highest correlation coefficient with dry weight of red pepper at the 40th, 50th, and 80th DAT, respectively. Also these reflectance indices at the same growth station was closely correlated with dry weight, yield, and nitrogen uptake of red pepper at the 120th DAT, especially showing the best correlation coefficient at the 80th DAT. From these result, the aNDVI at the 80th DAT can significantly explain for dry weight of red pepper at the 120th DAT as well as for application level of nitrogen fertilizer. Consequently ground remote sensing as a non-destructive real-time assessment of plant nitrogen status was thought to be a useful tool for in season nitrogen management for red pepper providing both spatial and temporal information.