• Journal of The Korean Society of Grassland and Forage Science
• /
• v.27 no.2
• /
• pp.93-100
• /
• 2007

This study was conducted to determine the effects of different levels (10, 12 and 15%) of crude protein (CP) in total mixed ration (TMR) on dry matter intake, digestibility and nitrogen balance of Korean black goats in the stage of early pregnancy and to obtain information on their optimal dietary levels of CP. In the present study, 12 Does of Korean black goats in the early pregnancy were allotted to four unreplicated groups by dietary level of CP and then they were housed in individual metabolism cages with completely randomized design throughout 30 days with 20 days adaptation and 10 days collection periods. Does in Control were fed a conventional diet and does in TMR10, TMR12 and TMR15 were fed a diet adjusted to about 10, 12 and 15% CP, respectively. Dry matter(DM) contents ranged from 89 to 91% in treatments. There were no differences fur fiber contents among three CP levels of TMR, showing that ADF and NDF had 18.57 to 19.85, and 53.41 to 54.80, respectively. Crude protein contents for three TMR treaements had 10.61, 12.15 and 14.97%, respectively. However, non-fibrous carbohydrate (NFC) contents decreased with increasing CP levels in treatments. Meanwhile, Intakes of DM, nutrients and digestible nutrients were significantly (p<0.05) higher in TMR15 and control than in TMR10 and TMR12. Moreover, DM intake per metabolic body weight and theit ratio per body weight was significantly (p<0.05) higher for control and TMR15 than other treatments. DM digestibility was not significantly different among treatments, but ether extract digestibility of treatments was significantly (p<0.05) higher than that of control, but there was no significant difference among treatments. Nitrogen retention significantly (p<0.05) increased with increasing CP levels in TMR, and TMR15 was highest among treatments. Our results showed that the increasing CP levels in TMR increased DM intake and nitrogen retention and suggested that the optimal dietary CP levels under TMR feeding system in early pregnant Korean black goats could be estimated for at least 15%.

• Journal of Oral Medicine and Pain
• /
• v.33 no.4
• /
• pp.339-352
• /
• 2008

Aims: The aims of this study were to identify the association between cervical muscle pain and TMD by pressure pain response, and to find cervical muscles showing moderate to severe pressure pain that are correlated with masticatory muscle pain. Methods: Patients(n=129, female 65.9%, mean age 28.8 years) answered a TMD questionnaire asking about headache, neck pain, emotional stress, sleep disturbance, parafunction habits, and pain intensity. A clinical examination of the masticatory system was performed. Of the neck muscles, (1) the upper sternocleidomastoid, (2) the middle sternocleidomastoid, (3) the upper trapezius, (4) the splenius capitis, (5) the semispinalis capitis, (6) the scalene medius, and (7) the levator scapulae muscles were examined by palpation. Pressure pain or tenderness of all palpation sites was scored from 0 to 3 according to the pain response. The variables of sum of pressure pain scores were calculated from pressure pain scores and were used for statistical analyses. Results: Eighty patients(62.0%) answered that they suffer from neck pain in the TMD questionnaire. More than 40% of sternocleidomastoid and upper trapezius examination sites showed moderate to severe tenderness in the cervical muscles, and 36% of middle masseter in the masticatory muscles. For the 129 patients, the sum of cervical muscle pain scores(mean=12.88, SD=8.06) and the sum of TMD pain scores(mean=5.36, SD=5.10) were moderately correlated($\rho$ = 0.502, P < 0.001). The sum of TMD pain scores tends to increase as the sum of cervical muscle pain scores increases(Y = 0.395${\cdot}$X, $R^2$ = 0.659, P < 0.001). In the patients with masticatory muscle disorders, the sum of sternocleidomastoid and upper trapezius pain scores(mean = 8.67, SD = 4.95) and the sum of temporalis and masseter pain scores(mean = 3.37, SD = 3.56) showed moderate correlation($\rho$ = 0.375, P < 0.001). Those two variables were in a proportionate relationship(Y = 0.359${\cdot}$X, $R^2$ = 0.538, P < 0.001). In a partial correlation analysis of the sum of unilateral pain scores, the sum of right cervical muscle pain scores and the sum of left cervical muscle pain scores showed the highest correlation(r = 0.802, P < 0.001). The sum of right TMD pain scores and the sum of left TMD pain scores were moderately correlated(r = 0.481, P < 0.001). For the twenty patients with unilateral TMD pain, the partial correlation coefficient between the sum of ipsilateral cervical muscle pain scores and the sum of contralateral cervical muscle pain scores was the largest(r = 0.597, P = 0.009). A partial correlation between the sum of primary TMD side pain scores and the sum of ipsilateral cervical muscle pain scores was 0.564(P = 0.015). Conclusions: TMD is associated with cervical muscle pain on condition of pressure pain response to palpation. Of the cervical muscles, sternocleidomastoid and upper trapezius frequently exhibit moderate to severe pressure pain, and they are closely related to the masticatory muscle pain. The characteristic of symmetric involvement of pain is prominent in cervical muscles; however, TMD can affect the level of cervical muscle pain to modify its symmetric nature.

• The Korean Journal of Air & Space Law and Policy
• /
• v.28 no.1
• /
• pp.115-157
• /
• 2013

Some contracting States of the Convention on International Civil Aviation (commonly known as the Chicago Convention) issue FAOC(Foreign AOC and/or Operations Specifications) and conduct various safety audits for the foreign operators. These FAOC and safety audits on the foreign operators are being expanded to other parts of the world. While this trend is the strengthening measure of aviation safety resulting in the reduction of aircraft accident, it is the source of concern from the legal as well as economic perspectives. FAOC of the USA doubly burdens the other contracting States to the Chicago Convention because it is the requirement other than that prescribed by the Chicago Convention of which provisions are faithfully observed by almost all the contracting States. The Chicago Convention in its Article 33 stipulates that each contracting State recognize the validity of the certificates of airworthiness and licenses issued by other contracting States as long as they meet the minimum standards of the ICAO. Consequently, it is submitted that the unilateral action of the USA, China, Mongolia, Australia, and the Philippines issuing the FOAC to the aircraft of other States is against the Convention. It is worry some that this breach of international law is likely to be followed by the European Union which is believed to be in preparation for its own unilateral application. The ICAO established by the Chicago Convention to be in charge of safe and orderly development of the international civil aviation has been in hard work to both upgrade and emphasize the safe operation of aircraft. As the result of these endeavors, it prepared a new Annex 19 to the Chicago Convention with the title of "Safety Management" and with the applicable date 14 November 2013. It is this Annex and other ICAO documents relevant to the safety that the contracting States to the Chicago Convention have to observe. Otherwise, it is the economical burden due to probable delay in issuing the FOAC and bureaucracies combined with many different paperworks and regulations depending on where the aircraft is flown. It is exactly to avoid this type of confusion and waste that the Chicago Convention aimed at when it was adopted in 1944. The State of the operator shall establish a system for both the certification and the continued surveillance of the operator in accordance with ICAO SARPs to ensure that the required standards of operations are maintained. Certainly the operator shall meet and maintain the requirements established by the States in which it operate. The authority of a State stops where the authority of another State intervenes or where the former has yielded its power by an international agreement for the sake of international cooperation. Hence, it is not within the realm of the State to issue FAOC towards foreign operators for the reason that these foreign operators are flying in and out of the State. Furthermore, there are other safety audits such as ICAO USOAP, IATA IOSA, FAA IASA, and EU SAFA that assure the safe operation of the aircraft, but within the limit of their power and in compliance with the ICAO SARPs. If the safety level of any operator is not satisfactory, the operator could be banned to operate in the contracting States with watchful eyes until the ICAO SARPs are met. This time-honoured practice has been applied without any serious problems. Besides, we have the new Annex 19 to strengthen and upgrade with easy reference for contracting States. We don't have no reason to introduce additional burden to the States by unilateral actions of some States. These actions have to be corrected. On the other hand, when it comes to the carriage of the Personal or Pilot Log Book, the Korean regulation requiring it is in contrast with other relevant provisions of USA, USOAP, IOSA, and SAFA. The Chicago Convention requires in its Articles 29 and 34 only the carriage of the Journey Log Book and some other certificates, but do not mention the Personal Log Book at all. Paragraph 5.1.1.1 of Annex 1 to the Chicago Convention even makes it clear that the carriage in the aircraft of the Personal Log Book is not required on international flights. The unique Korean regulation in this regards giving the unnecessary burden to the national flag air carriers has to be lifted at once.

• Korean journal of applied entomology
• /
• v.14 no.3 s.24
• /
• pp.97-109
• /
• 1975

In an attempt to develop an effective integrated system of controlling blast disease of rice caused by Pyricularia oryzae Cav., the possibility of minimizing the disease incidence by proper application of fertilizers has been investigated. Thus the effect of silicate, nitrogen, phosphorus and potassium fertilizers on the development of blast disease as well as the correlation between the rice varieties an4 strains of P. oryzae were studied. The experiments were made in 1971 and 1973 by artificial inoculation and under natural development of the blast disease on rice plants. The results obtained are summarized as follows. 1. Application of silicate fertilizer resulted in the increase of silicate as well as total sugar and potassium content but decrease of total nitrogen and phosphorus in tile leaf blades of rice plants. 2. The ratios of total C/total N. $ SiO_2/total$ N, and $K_2O/total$ N in leaf blades of rice plants increased by the application of silicate fertilizers. There was high level of negative correlation between the ratios mentioned above and the incidence of rice blast disease. 3. Application of silicate fertilizer reduced the incidence of rice blast disease. 4. The over dressing of nitrogen fertilizer resulted in the increase of total nitrogen and decrease of silicate and total sugar content in leaf blades, thus disposing the rice plants more susceptible to blast disease. 5. Over dressing of phosphorus fertilizer resulted in the increase of both total nitrogen and Phosphorus, and decrease of silicate content in the leaf blades inducing the rice plants to become more susceptible to blast disease. 6. Increased dressing of potash resulted in the increase of silicate content and $K_2O/total$ N ratio but decrease of total nitrogen content in leaf blades. When potassium content is low in the leaf blades of rice plants, the additional dressing of potash to rice plant contributed to the increase of resistance to blast disease. However, there was no significant correlation between additional potassium application and the resistance to blast disease when the potassium content is already high in the leaf blades. 7. When four rice varieties were artificially inoculated with three strains of P. oryzae, the incidence of blast disease was most severe on Pungok, least severe on Jinheung and moderate on Pungkwang and Paltal varieties. 8. Disease incidence was most severe on the second leaf from top and less sever on top and there leaf regardless of the fertilizer application when 5-6 leaf stage rice seedlings of four rice varieties were artificially inoculated with three strains of P. oryzae. 9. The pathogenicity of three strains of P. oryzae was in the order of $P_1,\;P_2,\;and\;P_3$ in their virulence when inoculated to Jinheung, Paltal, Pungkwang varieties but not with Pungok. The interaction between strains of P. oryzae and rice varieties was significant.

• Tuberculosis and Respiratory Diseases
• /
• v.55 no.3
• /
• pp.250-256
• /
• 2003

Background : Anthracofibrosis, a descriptive term for multiple black pigmentation with fibrosis on bronchoscopic examination, has a close relationship with active tuberculosis (TB). However, TB activity is determined in the later stage by the TB culture results in some cases of anthracofibrosis. Therefore, it is necessary to identify early markers of TB activity in anthracofibrosis. There have been several reports investigating the serum levels of IL-2 $sR{\alpha}$, IFN-${\gamma}$ and TBGL antibody for the evaluation of TB activity. In the present study, we tried to measure the above mentioned serologic markers for the evaluation of TB activity in patients with anthracofibrosis. Methods : Anthracofibrosis was defined when there was deep pigmentation (in more than two lobar bronchi) and fibrotic stenosis of the bronchi on bronchoscopic examination. The serum of patients with anthracofibrosis was collected and stored under refrigeration before the start of anti-TB medication. The serum of healthy volunteers (N=16), patients with active TB prior to (N=22), and after (N=13), 6 month-medication was also collected and stored. Serum IL-2 $sR{\alpha}$, IFN-${\gamma}$ were measured with ELISA kit (R&D system, USA) and serum TBGL antibody was measured with TBGL EIA kit (Kyowa Inc, Japan). Results : Serum levels of IL-2 $sR{\alpha}$ in healthy volunteers, active TB patients before and after medication, and patients with anthracofibrosis were $640{\pm}174$, $1,611{\pm}2,423$, $953{\pm}562$, and $863{\pm}401$ pg/ml, respectively. The Serum IFN-${\gamma}$ levels were 0, $8.16{\pm}17.34$, $0.70{\pm}2.53$, and $2.33{\pm}6.67$ pg/ml, and TBGL antibody levels were $0.83{\pm}0.80$, $5.91{\pm}6.71$, $6.86{\pm}6.85$, and $3.22{\pm}2.59$ U/ml, respectively. The serum level of TBGL antibody was lower than of other groups (p<0.05). There was no significant difference of serum IL-2 $sR{\alpha}$ and IFN-${\gamma}$ levels among the four groups. Conclusion : The serum levels of IL-2 $sR{\alpha}$, IFN-${\gamma}$ and TBGL antibody were not useful in the evaluation of TB activity in patients with anthracofibrosis. More useful ways need to be developed for the differentiation of active TB in patients with anthracofibrosis.

• Journal of Intelligence and Information Systems
• /
• v.23 no.3
• /
• pp.29-44
• /
• 2017

In Internet commerce, consumers are heavily influenced by product reviews written by other users who have already purchased the product. However, as the product reviews accumulate, it takes a lot of time and effort for consumers to individually check the massive number of product reviews. Moreover, product reviews that are written carelessly actually inconvenience consumers. Thus many online vendors provide mechanisms to identify reviews that customers perceive as most helpful (Cao et al. 2011; Mudambi and Schuff 2010). For example, some online retailers, such as Amazon.com and TripAdvisor, allow users to rate the helpfulness of each review, and use this feedback information to rank and re-order them. However, many reviews have only a few feedbacks or no feedback at all, thus making it hard to identify their helpfulness. Also, it takes time to accumulate feedbacks, thus the newly authored reviews do not have enough ones. For example, only 20% of the reviews in Amazon Review Dataset (Mcauley and Leskovec, 2013) have more than 5 reviews (Yan et al, 2014). The purpose of this study is to analyze the factors affecting the usefulness of online product reviews and to derive a forecasting model that selectively provides product reviews that can be helpful to consumers. In order to do this, we extracted the various linguistic, psychological, and perceptual elements included in product reviews by using text-mining techniques and identifying the determinants among these elements that affect the usability of product reviews. In particular, considering that the characteristics of the product reviews and determinants of usability for apparel products (which are experiential products) and electronic products (which are search goods) can differ, the characteristics of the product reviews were compared within each product group and the determinants were established for each. This study used 7,498 apparel product reviews and 106,962 electronic product reviews from Amazon.com. In order to understand a review text, we first extract linguistic and psychological characteristics from review texts such as a word count, the level of emotional tone and analytical thinking embedded in review text using widely adopted text analysis software LIWC (Linguistic Inquiry and Word Count). After then, we explore the descriptive statistics of review text for each category and statistically compare their differences using t-test. Lastly, we regression analysis using the data mining software RapidMiner to find out determinant factors. As a result of comparing and analyzing product review characteristics of electronic products and apparel products, it was found that reviewers used more words as well as longer sentences when writing product reviews for electronic products. As for the content characteristics of the product reviews, it was found that these reviews included many analytic words, carried more clout, and related to the cognitive processes (CogProc) more so than the apparel product reviews, in addition to including many words expressing negative emotions (NegEmo). On the other hand, the apparel product reviews included more personal, authentic, positive emotions (PosEmo) and perceptual processes (Percept) compared to the electronic product reviews. Next, we analyzed the determinants toward the usefulness of the product reviews between the two product groups. As a result, it was found that product reviews with high product ratings from reviewers in both product groups that were perceived as being useful contained a larger number of total words, many expressions involving perceptual processes, and fewer negative emotions. In addition, apparel product reviews with a large number of comparative expressions, a low expertise index, and concise content with fewer words in each sentence were perceived to be useful. In the case of electronic product reviews, those that were analytical with a high expertise index, along with containing many authentic expressions, cognitive processes, and positive emotions (PosEmo) were perceived to be useful. These findings are expected to help consumers effectively identify useful product reviews in the future.

• Journal of Intelligence and Information Systems
• /
• v.26 no.3
• /
• pp.51-69
• /
• 2020

Now, it is a stylized fact that a small number of technology firms such as Apple, Alphabet, Microsoft, Amazon, Facebook and a few others have become larger and dominant players in an industry. Coupled with the rise of these leading firms, we have also observed that a large number of young firms have become an acquisition target in their early IPO stages. This indeed results in a sharp decline in the number of new entries in public exchanges although a series of policy reforms have been promulgated to foster competition through an increase in new entries. Given the observed industry trend in recent decades, a number of studies have reported increased concentration in most developed countries. However, it is less understood as to what caused an increase in industry concentration. In this paper, we uncover the mechanisms by which industries have become concentrated over the last decades by tracing the changes in industry concentration associated with a firm's status change in its early IPO stages. To this end, we put emphasis on the case in which firms are acquired shortly after they went public. Especially, with the transition to digital-based economies, it is imperative for incumbent firms to adapt and keep pace with new ICT and related intelligent systems. For instance, after the acquisition of a young firm equipped with AI-based solutions, an incumbent firm may better respond to a change in customer taste and preference by integrating acquired AI solutions and analytics skills into multiple business processes. Accordingly, it is not unusual for young ICT firms become an attractive acquisition target. To examine the role of M&As involved with young firms in reshaping the level of industry concentration, we identify a firm's status in early post-IPO stages over the sample periods spanning from 1990 to 2016 as follows: i) being delisted, ii) being standalone firms and iii) being acquired. According to our analysis, firms that have conducted IPO since 2000s have been acquired by incumbent firms at a relatively quicker time than those that did IPO in previous generations. We also show a greater acquisition rate for IPO firms in the ICT sector compared with their counterparts in other sectors. Our results based on multinomial logit models suggest that a large number of IPO firms have been acquired in their early post-IPO lives despite their financial soundness. Specifically, we show that IPO firms are likely to be acquired rather than be delisted due to financial distress in early IPO stages when they are more profitable, more mature or less leveraged. For those IPO firms with venture capital backup have also become an acquisition target more frequently. As a larger number of firms are acquired shortly after their IPO, our results show increased concentration. While providing limited evidence on the impact of large incumbent firms in explaining the change in industry concentration, our results show that the large firms' effect on industry concentration are pronounced in the ICT sector. This result possibly captures the current trend that a few tech giants such as Alphabet, Apple and Facebook continue to increase their market share. In addition, compared with the acquisitions of non-ICT firms, the concentration impact of IPO firms in early stages becomes larger when ICT firms are acquired as a target. Our study makes new contributions. To our best knowledge, this is one of a few studies that link a firm's post-IPO status to associated changes in industry concentration. Although some studies have addressed concentration issues, their primary focus was on market power or proprietary software. Contrast to earlier studies, we are able to uncover the mechanism by which industries have become concentrated by placing emphasis on M&As involving young IPO firms. Interestingly, the concentration impact of IPO firm acquisitions are magnified when a large incumbent firms are involved as an acquirer. This leads us to infer the underlying reasons as to why industries have become more concentrated with a favor of large firms in recent decades. Overall, our study sheds new light on the literature by providing a plausible explanation as to why industries have become concentrated.

• Journal of Intelligence and Information Systems
• /
• v.20 no.1
• /
• pp.163-176
• /
• 2014

Social media is becoming the platform for users to communicate their activities, status, emotions, and experiences to other people. In recent years, microblogs, such as Twitter, have gained in popularity because of its ease of use, speed, and reach. Compared to a conventional web blog, a microblog lowers users' efforts and investment for content generation by recommending shorter posts. There has been a lot research into capturing the social phenomena and analyzing the chatter of microblogs. However, measuring television ratings has been given little attention so far. Currently, the most common method to measure TV ratings uses an electronic metering device installed in a small number of sampled households. Microblogs allow users to post short messages, share daily updates, and conveniently keep in touch. In a similar way, microblog users are interacting with each other while watching television or movies, or visiting a new place. In order to measure TV ratings, some features are significant during certain hours of the day, or days of the week, whereas these same features are meaningless during other time periods. Thus, the importance of features can change during the day, and a model capturing the time sensitive relevance is required to estimate TV ratings. Therefore, modeling time-related characteristics of features should be a key when measuring the TV ratings through microblogs. We show that capturing time-dependency of features in measuring TV ratings is vitally necessary for improving their accuracy. To explore the relationship between the content of microblogs and TV ratings, we collected Twitter data using the Get Search component of the Twitter REST API from January 2013 to October 2013. There are about 300 thousand posts in our data set for the experiment. After excluding data such as adverting or promoted tweets, we selected 149 thousand tweets for analysis. The number of tweets reaches its maximum level on the broadcasting day and increases rapidly around the broadcasting time. This result is stems from the characteristics of the public channel, which broadcasts the program at the predetermined time. From our analysis, we find that count-based features such as the number of tweets or retweets have a low correlation with TV ratings. This result implies that a simple tweet rate does not reflect the satisfaction or response to the TV programs. Content-based features extracted from the content of tweets have a relatively high correlation with TV ratings. Further, some emoticons or newly coined words that are not tagged in the morpheme extraction process have a strong relationship with TV ratings. We find that there is a time-dependency in the correlation of features between the before and after broadcasting time. Since the TV program is broadcast at the predetermined time regularly, users post tweets expressing their expectation for the program or disappointment over not being able to watch the program. The highly correlated features before the broadcast are different from the features after broadcasting. This result explains that the relevance of words with TV programs can change according to the time of the tweets. Among the 336 words that fulfill the minimum requirements for candidate features, 145 words have the highest correlation before the broadcasting time, whereas 68 words reach the highest correlation after broadcasting. Interestingly, some words that express the impossibility of watching the program show a high relevance, despite containing a negative meaning. Understanding the time-dependency of features can be helpful in improving the accuracy of TV ratings measurement. This research contributes a basis to estimate the response to or satisfaction with the broadcasted programs using the time dependency of words in Twitter chatter. More research is needed to refine the methodology for predicting or measuring TV ratings.

• 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.

• Journal of Intelligence and Information Systems
• /
• v.26 no.2
• /
• pp.79-104
• /
• 2020

Recently, as deep learning has attracted attention, the use of deep learning is being considered as a method for solving problems in various fields. In particular, deep learning is known to have excellent performance when applied to applying unstructured data such as text, sound and images, and many studies have proven its effectiveness. Owing to the remarkable development of text and image deep learning technology, interests in image captioning technology and its application is rapidly increasing. Image captioning is a technique that automatically generates relevant captions for a given image by handling both image comprehension and text generation simultaneously. In spite of the high entry barrier of image captioning that analysts should be able to process both image and text data, image captioning has established itself as one of the key fields in the A.I. research owing to its various applicability. In addition, many researches have been conducted to improve the performance of image captioning in various aspects. Recent researches attempt to create advanced captions that can not only describe an image accurately, but also convey the information contained in the image more sophisticatedly. Despite many recent efforts to improve the performance of image captioning, it is difficult to find any researches to interpret images from the perspective of domain experts in each field not from the perspective of the general public. Even for the same image, the part of interests may differ according to the professional field of the person who has encountered the image. Moreover, the way of interpreting and expressing the image also differs according to the level of expertise. The public tends to recognize the image from a holistic and general perspective, that is, from the perspective of identifying the image's constituent objects and their relationships. On the contrary, the domain experts tend to recognize the image by focusing on some specific elements necessary to interpret the given image based on their expertise. It implies that meaningful parts of an image are mutually different depending on viewers' perspective even for the same image. So, image captioning needs to implement this phenomenon. Therefore, in this study, we propose a method to generate captions specialized in each domain for the image by utilizing the expertise of experts in the corresponding domain. Specifically, after performing pre-training on a large amount of general data, the expertise in the field is transplanted through transfer-learning with a small amount of expertise data. However, simple adaption of transfer learning using expertise data may invoke another type of problems. Simultaneous learning with captions of various characteristics may invoke so-called 'inter-observation interference' problem, which make it difficult to perform pure learning of each characteristic point of view. For learning with vast amount of data, most of this interference is self-purified and has little impact on learning results. On the contrary, in the case of fine-tuning where learning is performed on a small amount of data, the impact of such interference on learning can be relatively large. To solve this problem, therefore, we propose a novel 'Character-Independent Transfer-learning' that performs transfer learning independently for each character. In order to confirm the feasibility of the proposed methodology, we performed experiments utilizing the results of pre-training on MSCOCO dataset which is comprised of 120,000 images and about 600,000 general captions. Additionally, according to the advice of an art therapist, about 300 pairs of 'image / expertise captions' were created, and the data was used for the experiments of expertise transplantation. As a result of the experiment, it was confirmed that the caption generated according to the proposed methodology generates captions from the perspective of implanted expertise whereas the caption generated through learning on general data contains a number of contents irrelevant to expertise interpretation. In this paper, we propose a novel approach of specialized image interpretation. To achieve this goal, we present a method to use transfer learning and generate captions specialized in the specific domain. In the future, by applying the proposed methodology to expertise transplant in various fields, we expected that many researches will be actively conducted to solve the problem of lack of expertise data and to improve performance of image captioning.