• 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.27 no.3
• /
• pp.231-252
• /
• 2021

Artificial intelligence (AI) is a key technology that will change the future the most. It affects the industry as a whole and daily life in various ways. As data availability increases, artificial intelligence finds an optimal solution and infers/predicts through self-learning. Research and investment related to automation that discovers and solves problems on its own are ongoing continuously. Automation of artificial intelligence has benefits such as cost reduction, minimization of human intervention and the difference of human capability. However, there are side effects, such as limiting the artificial intelligence's autonomy and erroneous results due to algorithmic bias. In the labor market, it raises the fear of job replacement. Prior studies on the utilization of artificial intelligence have shown that individuals do not necessarily use the information (or advice) it provides. Algorithm error is more sensitive than human error; so, people avoid algorithms after seeing errors, which is called "algorithm aversion." Recently, artificial intelligence has begun to be understood from the perspective of the augmentation of human intelligence. We have started to be interested in Human-AI collaboration rather than AI alone without human. A study of 1500 companies in various industries found that human-AI collaboration outperformed AI alone. In the medicine area, pathologist-deep learning collaboration dropped the pathologist cancer diagnosis error rate by 85%. Leading AI companies, such as IBM and Microsoft, are starting to adopt the direction of AI as augmented intelligence. Human-AI collaboration is emphasized in the decision-making process, because artificial intelligence is superior in analysis ability based on information. Intuition is a unique human capability so that human-AI collaboration can make optimal decisions. In an environment where change is getting faster and uncertainty increases, the need for artificial intelligence in decision-making will increase. In addition, active discussions are expected on approaches that utilize artificial intelligence for rational decision-making. This study investigates the impact of artificial intelligence on decision-making focuses on human-AI collaboration and the interaction between the decision maker personal traits and advisor type. The advisors were classified into three types: human, artificial intelligence, and human-AI collaboration. We investigated perceived usefulness of advice and the utilization of advice in decision making and whether the decision-maker's personal traits are influencing factors. Three hundred and eleven adult male and female experimenters conducted a task that predicts the age of faces in photos and the results showed that the advisor type does not directly affect the utilization of advice. The decision-maker utilizes it only when they believed advice can improve prediction performance. In the case of human-AI collaboration, decision-makers higher evaluated the perceived usefulness of advice, regardless of the decision maker's personal traits and the advice was more actively utilized. If the type of advisor was artificial intelligence alone, decision-makers who scored high in conscientiousness, high in extroversion, or low in neuroticism, high evaluated the perceived usefulness of the advice so they utilized advice actively. This study has academic significance in that it focuses on human-AI collaboration that the recent growing interest in artificial intelligence roles. It has expanded the relevant research area by considering the role of artificial intelligence as an advisor of decision-making and judgment research, and in aspects of practical significance, suggested views that companies should consider in order to enhance AI capability. To improve the effectiveness of AI-based systems, companies not only must introduce high-performance systems, but also need employees who properly understand digital information presented by AI, and can add non-digital information to make decisions. Moreover, to increase utilization in AI-based systems, task-oriented competencies, such as analytical skills and information technology capabilities, are important. in addition, it is expected that greater performance will be achieved if employee's personal traits are considered.

• The Korean Journal of Emergency Medical Services
• /
• v.15 no.3
• /
• pp.31-44
• /
• 2011

Purpose : The purpose of emergency medical services(EMS) is to save human lives and assure the completeness of the body in emergency situations. Those who have been qualified on medical practice to perform such treatment as there is the risk of human life and possibility of major physical and mental injuries that could result from the urgency of time and invasiveness inflicted upon the body. In the emergency medical activities, 119 emergency medical technicians mainly perform the task but they are not able to perform such task independently and they are mandatory to receive medical direction. The purpose of this study is to examine the recognition and request for medical direction by 119 emergency medical technicians in order to provide basic information on the development of medical direction program suitable to the characteristics of EMS as well as for the studies on EMS for the sake of efficient operation of pre-hospital EMS. Method : Questionnaire via e-mail was conducted during July 1-31, 2010 for 675 participants who are emergency medical technicians, nurses and other emergency crews in Gyeongbuk. The effective 171 responses were used for the final analysis. In regards to the emergency medical technicians' scope of responsibilities defined in Attached Form 14, Enforcement regulations of EMS, t-test analysis was conducted by using the means and standard deviation of the level of request for medical direction on the scope of responsibilities of Level 1 & Level 2 emergency medical technicians as the scale of medical direction request. The general characteristics, experience result, the reason for necessity, emergency medical technicians & medical director request level, medical direction method, the place of work of the medical director, feedback content and improvement plan request level were analyzed through frequency and percentage. The level of experience in medical direction and necessity were analyzed through ${\chi}^2$ test. Results : In regards to the medical direction experience per qualification, the experience was the highest with 53.3% for Level 1 emergency medical technicians and 80.3% responded that experience was helpful. As for the recognition on the necessity of medical direction, 71.3% responded as "necessary" and it turned out to be the highest of 76.9% in nurses. As for the reason for responding "necessary", the reason for reducing the risk and side-effects from EMS for patients was the largest(75.4%), and the reason of EMS delay due to the request of medical direction was the highest(71.4%) for the reason for responding "not necessary". In regards to the request level of the task scope of emergency medical technicians, injection of certain amount of solution during a state of shock was the highest($3.10{\pm}.96$) for Level 1 emergency rescuers, and the endotracheal intubation was the highest($3.12{\pm}1.03$) for nurses, and the sublingual administration of nitroglycerine(NTG) during chest pain was the highest($2.62{\pm}1.02$) for Level 2 emergency medical technicians, and regulation of heartbeat using AED was the highest($2.76{\pm}.99$) for other emergency crews. For the revitalization of medical direction, the improvement in the capability of EMS(78.9%) was requested from emergency crew, and the ability to evaluate the medical state of patient was the highest(80.1%) in the level of request for medical director. The prehospital and direct medical direction was the highest(60.8%) for medical direction method, and the emergency medical facility was the highest(52.0%) for the placement of medical director, and the evaluation of appropriateness of EMS was the highest(66.1%) for the feedback content, and the reinforcement of emergency crew(emergency medical technicians) personnel was the highest(69.0%) for the improvement plan. Conclusion : The medical direction is an important policy in the prehospital EMS activity because 119 emergency medical technicians agreed the necessity of medical direction and over 80% of those who experienced medical direction said it was helpful. In addition, the simulation training program using algorithm and case study through feedback are necessary in order to enhance the technical capability of ambulance teams on the item of professional EMS with high level of request in the task scope of emergency medical technicians, and recognition of medical direction is the essence of the EMS field. In regards to revitalizing medical direction, the improvement of the task performance capability of 119 emergency medical technicians and medical directors, reinforcement of emergency medical activity personnel, assurance of trust between emergency medical technicians and the emergency physician, and search for professional operation plan of medical direction center are needed to expand the direct medical direction method for possible treatment beforehand through the participation by medical director even at the step in which emergency situation report is received.

• Journal of Intelligence and Information Systems
• /
• v.21 no.4
• /
• pp.93-110
• /
• 2015

The recommender system is a system which recommends products to the customers who are likely to be interested in. Based on automated information filtering technology, various recommender systems have been developed. Collaborative filtering (CF), one of the most successful recommendation algorithms, has been applied in a number of different domains such as recommending Web pages, books, movies, music and products. But, it has been known that CF has a critical shortcoming. CF finds neighbors whose preferences are like those of the target customer and recommends products those customers have most liked. Thus, CF works properly only when there's a sufficient number of ratings on common product from customers. When there's a shortage of customer ratings, CF makes the formation of a neighborhood inaccurate, thereby resulting in poor recommendations. To improve the performance of CF based recommender systems, most of the related studies have been focused on the development of novel algorithms under the assumption of using a single profile, which is created from user's rating information for items, purchase transactions, or Web access logs. With the advent of big data, companies got to collect more data and to use a variety of information with big size. So, many companies recognize it very importantly to utilize big data because it makes companies to improve their competitiveness and to create new value. In particular, on the rise is the issue of utilizing personal big data in the recommender system. It is why personal big data facilitate more accurate identification of the preferences or behaviors of users. The proposed recommendation methodology is as follows: First, multimodal user profiles are created from personal big data in order to grasp the preferences and behavior of users from various viewpoints. We derive five user profiles based on the personal information such as rating, site preference, demographic, Internet usage, and topic in text. Next, the similarity between users is calculated based on the profiles and then neighbors of users are found from the results. One of three ensemble approaches is applied to calculate the similarity. Each ensemble approach uses the similarity of combined profile, the average similarity of each profile, and the weighted average similarity of each profile, respectively. Finally, the products that people among the neighborhood prefer most to are recommended to the target users. For the experiments, we used the demographic data and a very large volume of Web log transaction for 5,000 panel users of a company that is specialized to analyzing ranks of Web sites. R and SAS E-miner was used to implement the proposed recommender system and to conduct the topic analysis using the keyword search, respectively. To evaluate the recommendation performance, we used 60% of data for training and 40% of data for test. The 5-fold cross validation was also conducted to enhance the reliability of our experiments. A widely used combination metric called F1 metric that gives equal weight to both recall and precision was employed for our evaluation. As the results of evaluation, the proposed methodology achieved the significant improvement over the single profile based CF algorithm. In particular, the ensemble approach using weighted average similarity shows the highest performance. That is, the rate of improvement in F1 is 16.9 percent for the ensemble approach using weighted average similarity and 8.1 percent for the ensemble approach using average similarity of each profile. From these results, we conclude that the multimodal profile ensemble approach is a viable solution to the problems encountered when there's a shortage of customer ratings. This study has significance in suggesting what kind of information could we use to create profile in the environment of big data and how could we combine and utilize them effectively. However, our methodology should be further studied to consider for its real-world application. We need to compare the differences in recommendation accuracy by applying the proposed method to different recommendation algorithms and then to identify which combination of them would show the best performance.

• Journal of Global Scholars of Marketing Science
• /
• v.17 no.3
• /
• pp.1-19
• /
• 2007

1. Purpose Rural tourism is made by individuals with different characteristics, needs and wants. It is important to have information on the characteristics and preferences of the consumers of the different types of existing rural accommodation. The stud aims to identify the determinants of consumer preference by type of accommodations. 2. Methodology 2.1 Sample Data were collected from 1000 people by telephone survey with three-stage stratified random sampling in seven metropolitan areas in Korea. Respondents were chosen by sampling internal on telephone book published in 2006. We surveyed from four to ten-thirty 0'clock afternoon so as to systematic sampling considering respondents' life cycle. 2.2 Two-step cluster Analysis Our study is accomplished through the use of a two-step cluster method to classify the accommodation in a reduced number of groups, so that each group constitutes a type. This method had been suggested as appropriate in clustering large data sets with mixed attributes. The method is based on a distance measure that enables data with both continuous and categorical attributes to be clustered. This is derived from a probabilistic model in which the distance between two clusters in equivalent to the decrease in log-likelihood function as a result of merging. 2.3 Multinomial Logit Analysis The estimation of a Multionmial Logit model determines the characteristics of tourist who is most likely to opt for each type of accommodation. The Multinomial Logit model constitutes an appropriate framework to explore and explain choice process where the choice set consists of more than two alternatives. Due to its ease and quick estimation of parameters, the Multinomial Logit model has been used for many empirical studies of choice in tourism. 3. Findings The auto-clustering algorithm indicated that a five-cluster solution was the best model, because it minimized the BIC value and the change in them between adjacent numbers of clusters. The accommodation establishments can be classified into five types: Traditional House, Typical Farmhouse, Farmstay house for group Tour, Log Cabin for Family, and Log Cabin for Individuals. Group 1 (Traditional House) includes mainly the large accommodation establishments, i.e. those with ondoll style room providing meals and one shower room on family tourist, of original construction style house. Group 2 (Typical Farmhouse) encompasses accommodation establishments of Ondoll rooms and each bathroom providing meals. It includes, in other words, the tourist accommodations Known as "rural houses." Group 3 (Farmstay House for Group) has accommodation establishments of Ondoll rooms not providing meals and self cooking facilities, large room size over five persons. Group 4 (Log Cabin for Family) includes mainly the popular accommodation establishments, i.e. those with Ondoll style room with on shower room on family tourist, of western styled log house. While the accommodations in this group are not defined as regards type of construction, the group does include all the original Korean style construction, Finally, group 5 (Log Cabin for Individuals)includes those accommodations that are bedroom western styled wooden house with each bathroom. First Multinomial Logit model is estimated including all the explicative variables considered and taking accommodation group 2 as base alternative. The results show that the variables and the estimated values of the parameters for the model giving the probability of each of the five different types of accommodation available in rural tourism village in Korea, according to the socio-economic and trip related characteristics of the individuals. An initial observation of the analysis reveals that none of variables income, the number of journey, distance, and residential style of house is explicative in the choice of rural accommodation. The age and accompany variables are significant for accommodation establishment of group 1. The education and rural residential experience variables are significant for accommodation establishment of groups 4 and 5. The expenditure and marital status variables are significant for accommodation establishment of group 4. The gender and occupation variable are significant for accommodation establishment of group 3. The loyalty variable is significant for accommodation establishment of groups 3 and 4. The study indicates that significant differences exist among the individuals who choose each type of accommodation at a destination. From this investigation is evident that several profiles of tourists can be attracted by a rural destination according to the types of existing accommodations at this destination. Besides, the tourist profiles may be used as the basis for investment policy and promotion for each type of accommodation, making use in each case of the variables that indicate a greater likelihood of influencing the tourist choice of accommodation.

• Investigative Magnetic Resonance Imaging
• /
• v.5 no.1
• /
• pp.33-37
• /
• 2001

Purpose : The water exchange rate between bulk water and bound water is an important parameter in deciding the efficiency of paramagnetic contrast agents. In this study, we evaluated the water exchange rates of various Gd-chelates using oxygen-17 NMR technique. Material and Methods : The samples (Gd-DTPA, Gd-DTPA-BMA, Gd-DOTA, Gd-EOB-DTPA) were prepared by mixing 5% $^{17}O-enriched$ water (Isotech, USA). The pH of the samples was adjusted to physiological value [pH=7.0] by buffer solution. The variable temperature $^{17}O-NMR$ measurements were performed using Bruker-600 (14.1 T, 81.3 MHz) spectrometer. Bruker VT-1000 temperature control units were used to stabilize the temperature. The $^{17}O$ spin-spin relaxation times (T2) were measured using Carr-Purcell-Meiboom-Gill (CPMG)I pulse sequence with 24 echo trains. The variable temperature T2 relaxation data were then fitted into Solomon-Bloembergen equations using least square fit algorithm to estimate the water exchange times. Results : From the measured $^{17}O-NMR$ relaxation rates, the determined water exchange rates at 300K are $0.42{\;}{\mu}s$ for Gd-DTPA, $1.99{\;}{\mu}s$ for Gd-DTPA-BMA, $0.27{\;}{\mu}s$ for Gd-DOTA, and $0.11{\;}{\mu}s$ for Gd-EOB-DTPA. The Gd-DTPA-BMA showed slowest exchange whereas Gd-EOB-DTPA had fastest water exchange rate. In addition, it was found that the water exchange rates (${\tau}_m$) of all samples had exponential temperature dependence with different decay constant. Conclusion : $^{17}O-NMR$ relaxation rate measurements, when combined with variable temperature technique, provide a solid tool for studying water exchange rate, which is very important in investigating the detailed mechanism of relaxation enhancement effect of the paramagnetic contrast agents.

• Korean journal of food and cookery science
• /
• v.18 no.6
• /
• pp.644-654
• /
• 2002

Brown color characteristics and antioxidizing activities were investigated for Doenjang under different processing conditions. Doenjang A was prepared directly from Meju and saline solution whereas Doenjang B was Prepared after separating soy sauce by soaking for 45 days. Both Doenjangs were aged for up to 180 days. Antioxidizing activity was studied in relation to the brown color characteristics using fat-soluble extract and water-soluble extract of Doenjang. The intensity of brown color was higher in the water-soluble Doenjang extract than the fat-soluble Doenjang extract. In the UV-VIS scanning spectra, water-soluble Doenjang extracts showed significant changes as the aging proceeded, but fat-soluble Doenjang extract did not. Antioxidizing activity of fat-soluble Doenjang extract increased as the aging period extended; however, no significant difference was detected in the water-soluble extract. Overall, Doenjang A showed higher contents of amino acids, reducing sugar, brown color, and antioxidizing activity, and the antioxidizing ability was higher in water-soluble Doenjang extract rather than in the fat-soluble Doenjang extract.

• Transactions of the Korean Society of Mechanical Engineers B
• /
• v.20 no.7
• /
• pp.2386-2396
• /
• 1996

A numerical simulation has been performed for isothermal and reacting flows in an exisymmetric, bluff-body research combustor. The present formulation is based on the density-weighted averaged Navier-Stokes equations together with a k-epsilon. turbulence model and a modified eddy-breakup combustion model. The PISO algorithm is employed for solution of thel Navier-Stokes system. Comparison between measurements and predictions are made for a centerline axial velocities, location of stagnation points, strength of recirculation zone, and temperature profile. Even though the numerical simulation gives acceptable agreement with experimental data in many respects, the present model is defictient in predicting the recoveryt rate of a central near-wake region, the non-isotropic turbulence effects, and variation of turbulent Schmidt number. Several possible explanations for these discrepancies have been discussed.

• Parasites, Hosts and Diseases
• /
• v.30 no.2
• /
• pp.65-74
• /
• 1992

The tegumental ultrastructure of juvenile and adult Echinostoma cinetorchis (Trematoda: Echinostomatidae) was observed by scanning electron microscopy. Three-day (juvenile) and 16-day (adult) worms were harvested from rats (Sprague-Dawley) experimentally fed the metacercariae from the laboratory-infected fresh water snail, Hippeutis cantori. The worms were fifed with 2.5% glutaraldehyde, processed routinely, and observed by an ISI Korea DS-130 scanning electron microscope. The 3-day old juvenile worms were elongated and ventrally curved, with their ventral sucker near the anterior two-fifths of the body. The head crown was bearing 37∼38 collar spines arranged in a zigzag pattern. The lips of the oral and ventral suckers had 8 and 5 type II sensory papillae respectively, and bewteen the spines, a few type III papillae were observed. Tongue or spade-shape spines were distributed anteriorly to the ventral sucker, whereas peg-like spines were distributed posteriorly and became sparse toward the posterior body. The spines of the dorsal surface were similar to those of the ventral surface. The 16-day old adults were leaf-like, and their oral and ventral suckers were located very closely. Aspinous head crown, oral and ventral suckers had type II and type III sensory papillae, and numerous type I papillae were distributed on the tegument anterior to the ventral sucker. Scale-like spines, with broad base and round tip, were distributed densely on the tegument anterior to the ventral sucker but they became sparse posteriorly. At the dorsal surface, spines were observed at times only at the anterior body. The results showed that the tegument of E. cinetorchis is similar to that of other echinostomes, but differs in the number and arrangement of collar spines, shape and distribution of tegumenal spines, and type and distribution of sensory papillae.

• Journal of Intelligence and Information Systems
• /
• v.17 no.3
• /
• pp.131-146
• /
• 2011

It is becoming more essential than ever for office professionals to become competent in information collection/gathering and problem solving in today's global business society. In particular, office professionals do not only assist simple chores but are also forced to make decisions as quickly and efficiently as possible in problematic situations that can end in either profit or loss to their company. Since office professionals rely heavily on their tacit knowledge to solve problems that arise in everyday business situations, it is truly helpful and efficient to refer to similar business cases from the past and share or reuse such previous business knowledge for better performance results. Case-based reasoning(CBR) is a problem-solving method which utilizes previous similar cases to solve problems. Through CBR, the closest case to the current business situation can be searched and retrieved from the case or knowledge base and can be referred to for a new solution. This reduces the time and resources needed and increase success probability. The main purpose of this study is to design a system called COKRS(Case-based reasoning Office Knowledge Recommender System) and develop a prototype for it. COKRS manages cases and their meta data, accepts key words from the user and searches the casebase for the most similar past case to the input keyword, and communicates with users to collect information about the quality of the case provided and continuously apply the information to update values on the similarity table. Core concepts like system architecture, definition of a case, meta database, similarity table have been introduced, and also an algorithm to retrieve all similar cases from past work history has also been proposed. In this research, a case is best defined as a work experience in office administration. However, defining a case in office administration was not an easy task in reality. We surveyed 10 office professionals in order to get an idea of how to define a case in office administration and found out that in most cases any type of office work is to be recorded digitally and/or non-digitally. Therefore, we have defined a record or document case as for COKRS. Similarity table was composed of items of the result of job analysis for office professionals conducted in a previous research. Values between items of the similarity table were initially set to those from researchers' experiences and literature review. The results of this study could also be utilized in other areas of business for knowledge sharing wherever it is necessary and beneficial to share and learn from past experiences. We expect this research to be a reference for researchers and developers who are in this area or interested in office knowledge recommendation system based on CBR. Focus group interview(FGI) was conducted with ten administrative assistants carefully selected from various areas of business. They were given a chance to try out COKRS in an actual work setting and make some suggestions for future improvement. FGI has identified the user-interface for saving and searching cases for keywords as the most positive aspect of COKRS, and has identified the most urgently needed improvement as transforming tacit knowledge and knowhow into recorded documents more efficiently. Also, the focus group has mentioned that it is essential to secure enough support, encouragement, and reward from the company and promote positive attitude and atmosphere for knowledge sharing for everybody's benefit in the company.