• 정보처리학회논문지B
• /
• 제19B권1호
• /
• pp.31-36
• /
• 2012

최근 사용자의 흥미에 맞는 아이템이나 서비스를 추천해 주는 추천 시스템에 대한 관심이 높아지고 있다. 최근 종료된 Netflix 경연대회(Netflix Prize)가 이 분야에 대한 연구자들의 연구 의욕을 고취시켰고, 특히 협력적 여과(Collaborative Filtering) 방법은 아이템의 종류에 상관없이 적용 가능한 범용성 때문에 활발히 연구되고 있다. 본 논문은 강화 학습을 이용해서 추천 시스템의 협력적 여과 문제를 푸는 방법을 제안한다. 강화 학습을 통해, 영화 평점 데이터에서 각 사용자가 평점을 매긴 순서에 따른 평점 간의 연관 관계를 학습하고자 하였다. 이를 위해 협력적 여과문제를 마르코프 결정 과정(Markov Decision Process)로 수학적으로 모델링하였고, 강화 학습의 가장 대표적인 알고리즘인 Q-learning을 사용해서 평가의 순서의 연관 관계를 학습하였다. 그리고 실제로 평가의 순서가 평가에 미치는 영향이 있음을 실험을 통해서 검증하였다.

• 벤처창업연구
• /
• 제13권3호
• /
• pp.113-124
• /
• 2018

본 연구는 전문가와 일반인 관객의 평가가 영화 흥행성과에 미치는 영향을 신호 이론(Signaling Theory)과 정보 비대칭(Information Asymmetry) 논의를 기반으로 실증 분석하였다. 영화에 대한 평가가 영화 흥행 성과에 대한 기존 연구는 주로 전문가나 일반인 관객 중 한 주체의 평가에만 중점을 두어 이들의 효과에 대해 설명함으로써 신호의 효과성(Signaling effectiveness)를 검증하는 데에는 다소 미흡한 점이 있었다. 또한 시간이 지남에 따라 변화하는 영화 평가의 추이에 대해서는 분석이 제대로 이루어지고 있지 않아, 영화 평점의 신호 효과성을 깊이 있게 밝히는 데에는 제한적이었다. 따라서 본 연구는 1) 전문가 평가와 일반인 평가의 차이점과 2) 시간의 흐름에 따른 평가의 변화의 추이가 영화 성과에 영향을 미칠 것으로 보고 이들 간의 관계를 밝히고자 하였다. 이를 위하여, 영화진흥위원회와 네이버를 토대로 2003년부터 2012년까지 개봉했던 1,141개 한국 영화 데이터와 이들에 대한 평점을 수집하여 분석을 실시하였다. 실증 분석 결과, 영화 개봉 전 전문가의 평가는 영화 흥행 성적에 영향을 미치지 않으며, 개봉 후에는 일반인의 평가가 전문가의 평가보다 영화 성과에 긍정적인 영향을 미치는 것으로 나타났다. 또한 시간이 지남에 따라 영화에 대한 긍정적인 평가가 증가할수록 영화 흥행 성적은 향상되는 것을 보여주었다.

• 공학교육연구
• /
• 제19권3호
• /
• pp.35-43
• /
• 2016

The purpose of this study is to explore directions for improvement of teaching at college of engineering based on analysis of differences from course evaluation of students and instructors. Data was collected from 86 instructors' ratings on courses and their 3004 students' ratings on courses at college of engineering in a two-year, a three-year college and a University from 2010 to 2013. The results of the survey indicate significant differences in the statistics from the several questions between the instructors and the students as well as between the course in a two-year, a three-year college and in a University. First, instructors' self evaluation of the course is higher than students' satisfaction ratings of the course on the average. Instructors' self evaluation are high on the questions 'The subject was proper for the course', 'The course provided the latest theory and trend of the field', and 'Fairness and objectivity about the exams and the assignments'. Also, the difference between Instructors and students on the questions is significant in the statistics. The professor must make sure that students know well how to organize the course content and the method for feedback to test result and homework. Second, instructors have higher satisfaction ratings on the six questions and students have higher satisfaction ratings on the one question('Make students participate in the class effectively') at a two-year and a three-year college. However, students have higher satisfaction ratings on the three questions('Make students participate in the class effectively', 'Concern about students' learning process', and 'Use of E-learning and media equipments') and instructors have higher satisfaction ratings on the one question. It means instructors at a University feel pressure on a teaching and they are unsatisfied with their teaching skills. Third, the result of comparing six parts of the questions shows that students' satisfaction ratings are higher on 'Students participation' and 'Application of media equipments' parts whereas instructors' self evaluation are higher on 'Exams and assignments' part. Fourth, the question 'Make students participate in the class effectively' is significant in statistic based on comparison of instructors and students, and comparison of in a college and a University. Students' satisfaction ratings are higher than instructors' self evaluation.

• 한국의류학회지
• /
• 제25권3호
• /
• pp.560-568
• /
• 2001

The aim of the present study was to measure intra-individual consistency in clothing and fabric evaluation and to examine its relation to the ratings. A sample of 93 female and 97 male university students rated clothing of 4 styles of daytime wear and 2 fabrics on 15 pairs of polar adjectives twice in 7-days interval. Correlation coefficients between the two ratings for each subject, intra-individual consistency in the evaluation, ranged from -0.12 to 0.89 and mean coefficient was 0.63 of female and -0.01 to 0.78 and mean coefficient was 0.54 of male. Based on the coefficients, the subjects were classified into three groups: high, medium, and low intra-individual consistency. Analysis of variance of mean ratings by the three groups revealed that significant difference existed in 24% of female and 23% of male in 90 combinations of 6 clothing and 15 semantic differential scales. Female of subjects with high intra-individual consistency were most likely definite to evaluate clothing, whereas the ones with low were least. But male subjects were not definite. Mean correlation coefficients for style evaluation subscales of female was 0.39, but male was 0.44. Among the semantic differential scales, high stability in the two ratings was observed for the synthetic clothing evaluation. Correlation coefficients for each clothing obtained from the mean score of the subjects in each semantics differential scale were around 0.98, including that the mean scores of the subjects in each scale could yield excellent stability in clothing evaluation.

• 한국자동차공학회논문집
• /
• 제18권5호
• /
• pp.37-44
• /
• 2010

Impact noise is induced in a car when it is driven on a harsh road or over some bumps. This noise occurs with the very high level of sound, which affects passengers in some way or other. Although it is impossible to clearly remove such noise, it is necessary to research an improvement in sound quality for impact noise. A new sound metric for impact sound is presented. This metric is verified by comparison between mean subjective ratings and several sound metrics. In this paper, more objective attributes are considered, which the attributes are expressing the level and modulation of sound. Three sound metrics are employed to get impact sound indexes for each course by the method of multiple linear regressions. The indexes are verified by considering the correlation between the estimated values from the multiple linear regressions and the mean subjective ratings by evaluators. Also, the subjective ratings on the indexes are estimated for the case in which some parts of suspension system are changed. The estimated ratings represent more reasonable or acceptable ratings. Thus, such indexes can be used for modification of the parts of suspension system under considering a good sound quality.

• The Journal of Asian Finance, Economics and Business
• /
• 제6권2호
• /
• pp.25-32
• /
• 2019

In the Korean capital market, there are three credit rating agencies. Potential credit ratings based on credibility in the financial market are calculated independently for each rating agency. It often happens that despite the fact that the grades of the rating agencies are the same and have the same rating system, their actual ratings are different, even for the same firm. In such circumstances, investors may wonder why. In this study, we assume that the cause is the information environment in which the company operates. The credit ratings of rating agencies are mainly classified into bonds or commercial papers. The bonds are rated primarily for long-term of three years or more, and commercial papers specify ratings for less than one year. The information environment to be verified in this study was observed with a commercial paper. Under the assumption the larger the analyst following is, the more transparent is the information environment, we analyzed the influence of the number of analysts following on the degree to which ratings conflicted among credit rating agencies. The results of our analysis confirmed that opinion conflict among credit rating agencies is clearly reduced for companies with good information environments.

• 대한간호학회지
• /
• 제49권2호
• /
• pp.161-170
• /
• 2019

Background: Quality ratings could provide vital information to help people in choosing a nursing home. Purpose: This study investigated factors aligned with quality ratings of nursing homes. Methods: We employed a cross-sectional descriptive design to assess publicly available data on 1,354 nursing homes with 30 or more beds in the Republic of Korea. After excluding 289 nursing homes with no reported quality-evaluation ratings, we analyzed the 2015 data of 1,065 nursing homes. To prevent multicollinearity among independent variables, we carefully selected the final set of variables based on clinical and theoretical meaningfulness to direct nursing care. Quality, the ordinal outcome, was scored from 1 to 5 with a higher score indicating higher quality of the organization. We constructed a multivariate ordered logistic regression model. Results: Higher quality ratings of nursing homes was significantly related to the number of unoccupied beds (OR=0.99, p=.024), registered nurses (RNs) (OR=1.30, p=.003), qualified care workers (OR=1.03, p=.011), cognitive-improvement programs (OR=1.05, p=.024), and other programs for residents' activities (OR=1.09, p<.001). Conclusion: The number of RNs had the strongest influence on the publicly reported quality rating, while the rating of qualified care workers demonstrated little effect and that of nursing assistants had no effect. The number of RNs could be used as a crucial indicator for high-quality homes; more resident-engaging programs also demonstrated better quality of nursing home care.

• 한국정보시스템학회지:정보시스템연구
• /
• 제32권3호
• /
• pp.133-150
• /
• 2023

Purpose This study aims to develop classification models using a decision tree algorithm to identify core keywords and rules influencing online consumer review evaluations for the robot vacuum cleaner on Amazon.com. The difference from previous studies is that we analyze core keywords that affect the evaluation results by dividing the subjects that evaluate online consumer reviews into self-evaluation (star ratings) and peer evaluation (helpfulness votes). We investigate whether the core keywords influencing star ratings and helpfulness votes vary across different products and whether there is a similarity in the core keywords related to star ratings or helpfulness votes across all products. Design/methodology/approach We used random under-sampling to balance the dataset. We progressively removed independent variables based on decreasing importance through backwards elimination to evaluate the classification model's performance. As a result, we identified classification models that best predict star ratings and helpfulness votes for each product's online consumer reviews. Findings We have identified that the core keywords influencing self-evaluation and peer evaluation vary across different products, and even for the same model or features, the core keywords are not consistent. Therefore, companies' producers and marketing managers need to analyze the core keywords of each product to highlight the advantages and prepare customized strategies that compensate for the shortcomings.

• Management Science and Financial Engineering
• /
• 제9권2호
• /
• pp.13-20
• /
• 2003

For inferring criteria in a performance appraisal. linear programming (LP) has been utilized as an alternative to policy capturing (PC). Previous policy capturing LP (PCLP) studies were limited to the criteria of exact numerical ratings. However. under certain evaluation circumstances, a ranged rating with a lower and upper bound may be preferred over an exact numerical value. Therefore, this study introduces a new LP model that allows ranged ratings. A simple example is given to illustrate our model.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제8권8호
• /
• pp.2948-2966
• /
• 2014

To promote recommendation services through prediction quality, some privacy-preserving collaborative filtering solutions are proposed to make e-commerce parties collaborate on partitioned data. It is almost probable that two parties hold ratings for the same users and items simultaneously; however, existing two-party privacy-preserving collaborative filtering solutions do not cover such overlaps. Since rating values and rated items are confidential, overlapping ratings make privacy-preservation more challenging. This study examines how to estimate predictions privately based on partitioned data with overlapped entries between two e-commerce companies. We consider both user-based and item-based collaborative filtering approaches and propose novel privacy-preserving collaborative filtering schemes in this sense. We also evaluate our schemes using real movie dataset, and the empirical outcomes show that the parties can promote collaborative services using our schemes.