• 지능정보연구
• /
• 제18권3호
• /
• pp.185-202
• /
• 2012

오늘날 정보사회에서는 정보에 대한 가치를 인식하고, 이를 위한 정보의 활용과 수집이 중요해지고 있다. 얼굴 표정은 그림 하나가 수천개의 단어를 표현할 수 있듯이 수천 개의 정보를 지니고 있다. 이에 주목하여 최근 얼굴 표정을 통해 사람의 감정을 판단하여 지능형 서비스를 제공하기 위한 시도가 MIT Media Lab을 필두로 활발하게 이루어지고 있다. 전통적으로 기존 연구에서는 인공신경망, 중회귀분석 등의 기법을 통해 사람의 감정을 판단하는 연구가 이루어져 왔다. 하지만 중회귀모형은 예측 정확도가 떨어지고, 인공신경망은 성능은 뛰어나지만 기법 자체가 지닌 과적합화 문제로 인해 한계를 지닌다. 본 연구는 사람들의 자극에 대한 반응으로서 나타나는 얼굴 표정을 통해 감정을 추론해내는 지능형 모형을 개발하는 것을 목표로 한다. 기존 얼굴 표정을 통한 지능형 감정판단모형을 개선하기 위하여, Support Vector Regression(이하 SVR) 기법을 적용하는 새로운 모형을 제시한다. SVR은 기존 Support Vector Machine이 가진 뛰어난 예측 능력을 바탕으로, 회귀문제 영역을 해결하기 위해 확장된 것이다. 본 연구의 제안 모형의 목적은 사람의 얼굴 표정으로부터 쾌/불쾌 수준 그리고 몰입도를 판단할 수 있도록 설계되는 것이다. 모형 구축을 위해 사람들에게 적절한 자극영상을 제공했을 때 나타나는 얼굴 반응들을 수집했고, 이를 기반으로 얼굴 특징점을 도출 및 보정하였다. 이후 전처리 과정을 통해 통계적 유의변수를 추출 후 학습용과 검증용 데이터로 구분하여 SVR 모형을 통해 학습시키고, 평가되도록 하였다. 다수의 일반인들을 대상으로 수집된 실제 데이터셋을 기반으로 제안모형을 적용해 본 결과, 매우 우수한 예측 정확도를 보임을 확인할 수 있었다. 아울러, 중회귀분석이나 인공신경망 기법과 비교했을 때에도 본 연구에서 제안한 SVR 모형이 쾌/불쾌 수준 및 몰입도 모두에서 더 우수한 예측성과를 보임을 확인할 수 있었다. 이는 얼굴 표정에 기반한 감정판단모형으로서 SVR이 상당히 효과적인 수단이 될 수 있다는 점을 알 수 있었다.

• 지식경영연구
• /
• 제22권4호
• /
• pp.27-40
• /
• 2021

Click-Through Rate(CTR) 예측은 추천시스템에서 후보 항목의 순위를 결정하고 높은 순위의 항목들을 추천하여 고객의 정보 과부하를 줄임과 동시에 판매 촉진을 통한 수익 극대화를 달성할 수 있는 핵심 기능이다. 자연어 처리와 이미지 분류 분야는 심층신경망(deep neural network)의 활용을 통한 괄목한 성장을 하고 있다. 최근 이 분야의 주류를 이루던 모델과 차별화된 어텐션(attention) 메커니즘 기반의 트랜스포머(transformer) 모델이 제안되어 state-of-the-art를 달성하였다. 본 연구에서는 CTR 예측을 위한 트랜스포머 기반 모델의 성능 향상 방안을 제시한다. 자연어와 이미지 데이터와는 다른 이산적(discrete)이며 범주적(categorical)인 CTR 데이터 특성이 모델 성능에 미치는 영향력을 분석하기 위해 임베딩의 일반화(regularization)와 트랜스포머의 정규화(normalization)에 관한 실험을 수행한다. 실험 결과에 따르면, CTR 데이터 입력 처리를 위한 임베딩 과정에서 L2 일반화의 적용과 트랜스포머 모델의 기본 정규화 방법인 레이어 정규화 대신 배치 정규화를 적용할 때 예측 성능이 크게 향상됨을 확인하였다.

• Journal of Mechanical Science and Technology
• /
• 제19권8호
• /
• pp.1670-1681
• /
• 2005

In-situ diagnosis of chiller performance is an essential step for energy saving business. The main purpose of the in-situ diagnosis is to predict the performance of a target chiller. Many models based on thermodynamics have been proposed for the purpose. However, they have to be modified from chiller to chiller and require profound knowledge of thermodynamics and heat transfer. This study focuses on developing an easy-to-use diagnostic technique that is based on adaptive neuro-fuzzy inference system (ANFIS). The effect of sample data distribution on training the ANFIS is investigated. It is found that the data sampling over 10 days during summer results in a reliable ANFIS whose performance prediction error is within measurement errors. The reliable ANFIS makes it possible to prepare an energy audit and suggest an energy saving plan based on the diagnosed chilled water supply system.

• 경영정보학연구
• /
• 제8권2호
• /
• pp.139-154
• /
• 2006

추천시스템은 사용자들의 관심을 끄는 아이템을 그들이 보다 쉽게 찾도록 도와주거나, 그들의 기호에 기반하여 의미 있는 아이템들을 제공한다. 지금까지 가장 성공적이었던 협업 필터링 기반 추천시스템은 다른 사용자들의 의견을 참조하여 추천을 원하는 사용자에게 추천을 한다. 즉, 아이템들에 대한 사용자 기호를 나타내는 다른 사용자들의 평가정보가 추천을 위한 정보원으로 사용된다. 이처럼, 협업 필터링 기반 추천시스템이 사용자들의 기호만을 이용하도록 설계되었지만, 다른 정보를 이용하면 추천시스템의 성능과 정확도를 높일 수 있을 것으로 사료되어, 본 논문에서는 유사 정도와 인구통계학 정보를 이용한 협업 필터링 기반 추천시스템을 제안한다. 이런 추천시스템에서는 평가정보가 계속적으로 누적되기 때문에, 추천시스템의 정확도를 유지할 수 있는 한, 사용하는 데이터의 양을 줄이는 게 중요하다. 본 논문에서는 유사 정도와 인구통계학 정보를, 사용할 데이터의 양을 줄이기 위한 기준으로 사용하여 자연스레 시스템의 성능을 향상시켰다. 본 논문에서는 실험을 통하여 유사 정도의 사용이 추천시스템의 정확도를 높여주었고, 특정 인구통계학 정보의 사용도 추천시스템의 정확도를 높였음을 보였다.

• 대한안전경영과학회지
• /
• 제10권2호
• /
• pp.71-83
• /
• 2008

Most researches on the corporate credit rating are generally classified into the area of bankruptcy prediction and bond rating. The studies on bankruptcy prediction have focused on improving the performance in binary classification problem, since the criterion variable is categorical, bankrupt or non-bankrupt. The other studies on bond rating have predicted the credit ratings, which was already evaluated by bond rating experts. The financial institute, however, should perform effective loan evaluation and risk management by employing the corporate credit rating model, which is able to determine the credit of corporations. Therefore, in this study we present a medium sized corporate credit rating system by using Artificial Neural Network(ANN) and Analytical Hierarchy Process(AHP). Also, we developed AHP model for credit rating using non-financial information. For the purpose of completed credit rating model, we integrated the ANN and AHP model using both financial information and non-financial information. Finally, the credit ratings of each firm are assigned by the proposed method.

• 품질경영학회지
• /
• 제50권4호
• /
• pp.701-716
• /
• 2022

Purpose: This study aims to predict the dry cargo transportation market economy. The subject of this study is the BDI (Baltic Dry Index) time-series, an index representing the dry cargo transport market. Methods: In order to increase the accuracy of the BDI time-series, we have pre-processed the original time-series via time-series decomposition and data augmentation techniques and have used them for ANN learning. The ANN algorithms used are Multi-Layer Perceptron (MLP), Recurrent Neural Network (RNN), and Long Short-Term Memory (LSTM) to compare and analyze the case of learning and predicting by applying time-series decomposition and data augmentation techniques. The forecast period aims to make short-term predictions at the time of t+1. The period to be studied is from '22. 01. 07 to '22. 08. 26. Results: Only for the case of the MAPE (Mean Absolute Percentage Error) indicator, all ANN models used in the research has resulted in higher accuracy (1.422% on average) in multivariate prediction. Although it is not a remarkable improvement in prediction accuracy compared to uni-variate prediction results, it can be said that the improvement in ANN prediction performance has been achieved by utilizing time-series decomposition and data augmentation techniques that were significant and targeted throughout this study. Conclusion: Nevertheless, due to the nature of ANN, additional performance improvements can be expected according to the adjustment of the hyper-parameter. Therefore, it is necessary to try various applications of multiple learning algorithms and ANN optimization techniques. Such an approach would help solve problems with a small number of available data, such as the rapidly changing business environment or the current shipping market.

• 한국정밀공학회지
• /
• 제33권5호
• /
• pp.401-407
• /
• 2016

This paper describes the derivation methodology of the working torque predictive model that can be used in the initial design stages of the impact hammer tool. The working torque control mechanism is designed, taking into account various factors, such as the force of the spring and friction. Firstly, the analysis dynamic model for working environments was modeled as an additional bush and spring, and verified by comparing the test results of the working torque. Secondly, the main performance parameters of the working torque were theoretically defined by analyzing the operating mechanism. The equation to predict the working torque was derived using the dynamic analysis results according to the value changes of the parameters. The prediction equation of the working torque was validated by comparing the predicted results with the experimental data. The error difference between the experimental data and the predictive model results was found to be 8.62%.

• The Journal of Asian Finance, Economics and Business
• /
• 제8권1호
• /
• pp.179-185
• /
• 2021

The aim of this study is to examine the impact of board of directors' activity on the audit committee's effectiveness score among manufactured listed companies on Saudi Stock Exchange (Tadawul) for the period 2015-2017. The final sample of this study consists of 195 firm-year observations that represent manufactured companies listed on Saudi Stock Exchange (Tadawul) for the years 2015-2017. The data of this study in terms of board of directors' meetings, audit committee size and meetings, firm leverage, firm performance, and firm age were hand-collected from the annual reports of the considered companies. The Pooled OLS regression's result indicate that audit committee's effectiveness score is influenced by the board of directors' activity. This result gives support to the agency theory prediction. This result is also consistent with the complementary function of corporate governance mechanisms in which board of directors' activity complements the function of audit committee's effectiveness score. The result of this study should be useful for manufacturing companies, Saudi Stock Exchange, auditors, and regulators which relates to the association between board of directors' activity and audit committee's effectiveness score. This study provides a new empirical evidence on the impact of board activity on the audit committee's effectiveness score in an interesting context which is Saudi Arabia.

• Journal of Korea Trade
• /
• 제26권8호
• /
• pp.1-20
• /
• 2022

Purpose - The paper aims to develop the theory of TAM and perceived risk through a more comprehensive and rigorous understanding of the influencing factors of the consumer's adoption of LG HomNet smart home from the perspective of trade-offs. Design/methodology - Based on the TAM and perceived risk theory, combined with the individual characteristics of consumers in the context of information technology as the external factors of the technology acceptance model, this paper constructs a theoretical model of the factors affecting the use intention of the consumer. It was empirically tested by using SEM, and survey data was collected from 458 respondents. Findings - The research results show that 9 hypotheses of the research model are supported and have reliable prediction accuracy. Consumers' perceived interest, perceived connectivity and perceived controllability have a significant positive impact on their intention to use. In addition, this paper also confirmed the mediating effect of perceived usefulness and perceived ease of use. Originality/value - Consumers are very concerned about gains and losses. Low-level performance risks, security risks, and financial risks will drive the consumer to have a stronger intention to use, and financial risks have the strongest impact. This research provides a useful implication and guidance for smart home equipment manufacturers and service providers in product and service innovation and marketing and promotion strategies.

• 대한산업공학회지
• /
• 제28권2호
• /
• pp.193-200
• /
• 2002

A study was conducted to investigate the availability of Fitts' Law to Hangul input systems on mobile phones. Three different Hangul input systems were experimented to measure the performance time to evaluate the physical interface of all. The measured performance time was found to be well fitted with the modified Fitts' Law by Hangul input systems on mobile phones. As a result, the physical interfaces for Hangul input systems could be evaluated quantitatively with the prediction of the performance time by Fitts' Law.