• 한국콘텐츠학회논문지
• /
• 제20권3호
• /
• pp.267-274
• /
• 2020

클릭률(CTR) 예측은 사용자가 주어진 항목을 클릭할 확률을 추정하는 것으로 온라인 광고 수익 극대화를 위한 전략 결정에 중요한 역할을 한다. 최근 CTR 예측을 위해 CNN을 활용하는 시도가 이루어지고 있다. CTR 데이터는 특징 정보가 연관성 측면에서 의미 있는 순서를 갖지 않기 때문에, 임의의 순서로 배열될 수 있다. 하지만 CNN은 필터 사이즈에 의해 제한된 로컬 정보만을 학습하기 때문에 데이터 배열이 성능에 큰 영향을 줄 수 있다. 이 논문에서는 CNN이 수집할 수 있는 모든 로컬 특징 정보를 추출할 수 있는 데이터 배열 집합을 생성하고 생성된 배열들에 대하여 개별 CNN 모듈들이 특징들을 학습할 수 있는 다중 배열 CNN 모델을 제안한다. 대규모 데이터 세트에 대한 실험 결과에 따르면 제안된 모델은 기존 CNN 대비 AUC의 RI에서 22.6% 상승 효과를, 제안된 배열 생성 방법은 임의 생성 방법보다 3.87% 성능 향상을 달성하였다.

• 한국디지털정책학회:학술대회논문집
• /
• 한국디지털정책학회 2005년도 추계학술대회
• /
• pp.45-60
• /
• 2005

The purpose of this study is to classify customer by e-mailing responsiveness on time-series analysis and testify the effectiveness of grouping by ROI analysis. Response recency, response frequency and Activity(RFA) of e-mailing systems were adapted for Customer segmentations. ROI analysis were consisted of open, click-through, duration time, personalization, conversion rate and email loyalty index of email systems. Major findings are as follows: RFA analysis is used for customer segmentations that is fundamental process of e-CRM applications. Customer segmentations were loyal customer, odds customer, dormant customer, secession customer and observation customer by RFA grouping. Loyal customer group has high point in all ROI index compared to other groups. These results indicated that customer responsiveness of e-mailing systems were appropriate methods to grouping the customer with demographic variables. Therefore, effective e-mailing marketing strategy of e-Biz have suitable active DB and Behavior targeting is best approach to enforcing the target e-mailing marketing.

• 기술혁신학회지
• /
• 제7권3호
• /
• pp.681-709
• /
• 2004

The purpose of this study is to classify customers by e-mail responsiveness on time-series analysis and testify the effectiveness of grouping by ROI analysis. Response recency, response frequency and Activity(RFA) of e-mailing systems are adapted for Customer segmentations. ROI analysis are consisted of open, click-through, duration time, personalization, conversion rate and email loyalty index of email systems. Major findings are as follows: RFA analysis is used for customer segmentations that is fundamental process of e-CRM applications. Customers can be grouped into loyal customers, odds customers, dormant customers, secession customers, and observation customers by RFA grouping. Loyal customer group has high point in all ROI index compared to other groups. These results indicated that customer responsiveness of e-mail systems were appropriate methods to group the customer with demographic variables. Therefore, effective e-mail marketing strategy of e-Biz should have suitable active DB and Behavior targeting is best approach to enforce the target e-mail marketing.

• Journal of Information Processing Systems
• /
• 제13권5호
• /
• pp.1052-1070
• /
• 2017

For the autoencoder (AE) implemented as a construction component, this paper uses the method of greedy layer-by-layer pre-training without supervision to construct the stacked autoencoder (SAE) to extract the abstract features of the original input data, which is regarded as the input of the logistic regression (LR) model, after which the click-through rate (CTR) of the user to the advertisement under the contextual environment can be obtained. These experiments show that, compared with the usual logistic regression model and support vector regression model used in the field of predicting the advertising CTR in the industry, the SAE-LR model has a relatively large promotion in the AUC value. Based on the improvement of accuracy of advertising CTR prediction, the enterprises can accurately understand and have cognition for the needs of their customers, which promotes the multi-path development with high efficiency and low cost under the condition of internet finance.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2023년도 추계학술발표대회
• /
• pp.491-492
• /
• 2023

클릭률(CTR) 예측은 많은 실제 응용 프로그램에서 가장 기본적인 작업 중 하나가 되었으며 이 분야에서 많은 고급 모델이 나았다. 그러나 가장 고전적인 CF(Collaborative Filtering) 모델에서 딥러닝 모델로 발전하는 과정에서 특징 교차의 기본 단위가 요소(비트 단위)가 아닌 특징(벡터 단위)이라는 사실을 기억하는 모델은 거의 없다. 이 논문에서는 Masknet 모델에 벡터 단위 교차를 적용하는 클릭률 예측 모델은 제안한다.Movielens 에 대해 예측 결과는 89.24%로 나타나고 원본 모델보다 효과가 더 좋다.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2019년도 추계학술발표대회
• /
• pp.781-784
• /
• 2019

최근 AI를 산업 서비스에 적용하기 위해 많은 회사들이 활발히 연구를 하고 있다. 아마존과 넷플릭스 같은 거대 기업들은 이미 빅데이터와 AI 머신러닝을 이용한 추천 시스템을 구현하였고 아마존은 매출의 35%가 추천에 의해 발생하고 넷플릭스 75%의 사용자가 추천을 통해 영화를 선택한다고 보고되었다. 이러한 두 기업의 높은 추천 효율성의 이유는 협업 필터링(Collaborative filtering)과 같은 다양한 추천 알고리즘과 방대한 상품 및 고객 행동(구매, 시청 등) 데이터 등이 존재하고 있기 때문이다. 기계학습에서 알고리즘 학습을 위한 데이터의 양이 많지 않을 경우 알고리즘의 성능을 보장할 수 없다는 것이 일반적인 의견이다. 방대한 데이터를 가진 기업에서 추천 알고리즘을 적극적으로 활용 및 연구하고 있는 것도 이러한 이유 때문이다. 반면, 오프라인 및 여행사 기반에서 온라인 기반으로 영역을 차츰 확대하고 있는 항공 서비스 고객 데이터의 경우, 산업의 특성상 많은 회원에 비해 고객 1명당 온라인에서 활동하는 이력이 많지 않은 것이 특징이다. 이는, 추천 알고리즘을 통한 서비스 제공에서 큰 제약사항으로 작용한다. 본 연구에서는, 이러한 희박한 고객 활동 데이터에서 최신성 기반의 추천 시스템을 통하여 제약사항을 극복하고 추천 효율을 높이는 방법을 제안한다. 고객의 최근 접속 이력 로그를 시간 기준으로 데이터 셋을 분할하여 추천 알고리즘에 반영하였을 때, 추천된 노선에 대한 고객의 반응을 추천 성능 지표인 CTR(Click-Through Rate)로 측정하여 성능을 확인해 보았다.

• 한국멀티미디어학회논문지
• /
• 제24권7호
• /
• pp.903-912
• /
• 2021

Eye tracking research for upper limb disabilities is showing an effect in the aspect of device control. However, the reality is that it is not enough to perform web interaction with only eye tracking technology. In the Eye-Voice interface, a previous study, in order to solve the problem that the existing gaze tracking interfaces cause a malfunction of pointer execution, a gaze tracking interface supplemented with a voice command was proposed. In addition, the reduction of the malfunction rate of the pointer was confirmed through a comparison experiment with the existing interface. In this process, the difficulty of pointing due to the small size of the execution object in the web environment was identified as another important problem of malfunction. In this study, we propose an auto-magnification interface of objects so that people with upper extremities can freely click web contents by improving the problem that it was difficult to point and execute due to the high density of execution objects and their arrangements in web pages.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제18권3호
• /
• pp.704-719
• /
• 2024

Botnet pandemics are becoming more prevalent with the growing use of mobile phone technologies. Mobile phone technologies provide a wide range of applications, including entertainment, commerce, education, and finance. In addition, botnet refers to the collection of compromised devices managed by a botmaster and engaging with each other via a command server to initiate an attack including phishing email, ad-click fraud, blockchain, and much more. As the number of botnet attacks rises, detecting harmful activities is becoming more challenging in handheld devices. Therefore, it is crucial to evaluate mobile botnet assaults to find the security vulnerabilities that occur through coordinated command servers causing major financial and ethical harm. For this purpose, we propose a hybrid analysis approach that integrates permissions and API and experiments on the machine-learning classifiers to detect mobile botnet applications. In this paper, the experiment employed benign, botnet, and malware applications for validation of the performance and accuracy of classifiers. The results conclude that a classifier model based on a simple decision tree obtained 99% accuracy with a low 0.003 false-positive rate than other machine learning classifiers for botnet applications detection. As an outcome of this paper, a hybrid approach enhances the accuracy of mobile botnet detection as compared to static and dynamic features when both are taken separately.

• 한국초등과학교육학회지:초등과학교육
• /
• 제32권3호
• /
• pp.327-345
• /
• 2013

This study aims to investigate how pre-service elementary school teachers understand 'something scientific', 'being scientific', 'scientific events' and 'scientific questions' through semantic network analysis. To achieve this purpose, this study carried out a central analysis of the frequency and density of words and the degree of connection between key words, a concentric analysis, a click analysis and a common network analysis through text semantic network analysis by using NetMiner 4.0 Program. Based on the results of these analyses, this study came to the following conclusions. Firstly, in perceiving 'something scientific', pre-service elementary school teachers recognized 'verification', 'objective' and 'experiment' as most important words. In other words, they perceived that main grounds for something scientific should be provided through clear facts, possible to be verified and accompanied by an exact and logical theoretical system. In regard to 'being scientific', they perceived 'explanation', 'objective' and 'verification' as most important words, while having a traditional point of view that science is a set that can be explained objectively. Secondly, in regard that the term, 'observation', is contained in 'scientific events', they showed a high rate of understanding it as a scientific event. In regard to scientifical reasons, they showed the highest frequency of 'observation', and for unscientific reasons, they showed the highest frequency of 'behavior'. In perceiving 'scientific questions', they showed the highest frequency of determining bacteria-related questions as scientific. As a reason why they thought as scientific, they mentioned 'observation' most frequently like 'scientific events', while mentioning 'value judgement' as a reason why they thought as unscientific most frequently. From the results of integrated network analysis, this study found out that words pre-service teachers commonly used in stating scientific events or scientific questions were overlapped with words they mentioned for scientific events or scientific questions. As a result, it was found there were many pre-service teachers having interpreted scientific words without clearly distinguishing scientific events or scientific questions.

• 대한인간공학회지
• /
• 제27권4호
• /
• pp.109-117
• /
• 2008

Compared to the existing PC which uses a mouse and a keyboard, the touchscreen-based portable PC allows the user to use fingers, requiring new operation methods. However, current touchscreen-based web browser operations in many cases involve merely having fingers move simply like a mouse and click, or not corresponding well to the user's sensitivity and the structure of one's index finger, making itself difficult to be used during walking. Therefore, the goal of this study is to develop finger gestures which facilitate the interaction between the interface and the user, and make the operation easier. First, based on the frequency of usage in the web browser and preference, top eight functions were extracted. Then, the users' structural knowledge was visualized through sketch maps, and the finger gestures which were applicable in touchscreens were derived through the Meaning in Mediated Action method. For the front/back page, and up/down scroll functions, directional gestures were derived, and for the window closure, refresh, home and print functions, letter-type and icon-type gestures were drawn. A validation experiment was performed to compare the performance between existing operation methods and the proposed one in terms of execution time, error rate, and preference, and as a result, directional gestures and letter-type gestures showed better performance than the existing methods. These results suggest that not only during the operation of touchscreen-based web browser in portable PC but also during the operation of telematics-related functions in automobile, PDA and so on, the new gestures can be used to make operation easier and faster.