• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.05a
• /
• pp.657-659
• /
• 2023

최근 스마트폰과 같은 디지털 기기의 보급과 함께 개인화, 맞춤형 서비스의 수요가 늘어나면서 추천 서비스가 주목을 받고 있다. 세션 기반(Session based) 추천 시스템은 사용자의 아이템 선호에 따른 순서 정보를 고려한 학습 추천 모델로, 다양한 산업 분야에서 사용되고 있다. 세션 기반 추천 시스템 중 SASRec(Self-Attentive Sequential Recommendation) 모델은 MC/CNN/RNN 기반의 기존 여러 순차 모델들에 비하여 효율적인 성능을 보인다. 본 연구에서는 SASRec 모델의 하이퍼파라미터 중 배치 사이즈(Batch Size), 학습률 (Learning Rate), 히든 유닛(Hidden Unit)을 조정하여 실험함으로써 하이퍼파라미터에 의한 성능 변화를 분석하였다.

• Journal of Convergence for Information Technology
• /
• v.9 no.6
• /
• pp.83-90
• /
• 2019

People constantly strive for better choices. For this reason, recommender system has been developed since the early 1990s. In particular, collaborative filtering technique has shown excellent performance in the field of recommender systems, and research of recommender system using machine learning has been actively conducted. This study constructs recommender system using collaborative filtering and machine learning based on stacking model which is one of ensemble methods. The results of this study confirm that the recommender system with the stacking model is useful in aspects of recommender performance. In the future, the model proposed in this study is expected to help individuals or firms to make better choices.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.12
• /
• pp.1809-1816
• /
• 2021

In order to identify the latest research trends using data related to national R&D projects and to produce and utilize meaningful information, the application of automatic classification technology was also required in the national R&D information service, so we conducted research to automatically classify and recommend research field. About 450,000 cases of national R&D project data from 2013 to 2020 were collected and used for learning and evaluation. A model was selected after data pre-processing, analysis, and performance analysis for valid data among collected data. The performance of Word2vec, GloVe, and fastText was compared for the purpose of deriving the optimal model combination. As a result of the experiment, the accuracy of only the subcategories used as essential items of task information is 90.11%. This model is expected to be applicable to the automatic classification study of other classification systems with a hierarchical structure similar to that of the national science and technology standard classification research field.

• The Journal of Korean Association of Computer Education
• /
• v.14 no.4
• /
• pp.63-70
• /
• 2011

Recommender systems, which predict and recommend items that may possibly draw users' interests, have been applied in various fields as e-commerce systems are widespread. Collaborative filtering, one of the major methodologies of recommender systems, recommends either items similar to those preferred by the user, or items preferred by the other similar user. Therefore, two problems determine its performance; one is correct estimation of similarity and the other is predicting the real rating of the recommended item. This study addresses the latter problem. Previous studies predict the real rating based on the mean of the ratings, but this study proposes a prediction based on the range of the ratings and investigates its performance through experiments. As a result, it is demonstrated that the proposed method improves the mean absolute error significantly, compared to the previous method.

• Journal of the Korean Society for information Management
• /
• v.24 no.2
• /
• pp.89-104
• /
• 2007

This research attempts to give a personalized recommendation framework in large-sized music contents environment. Despite of existing studios and commercial contents for recommendation systems, large online shopping malls are still looking for a recommendation system that can serve personalized recommendation and handle large data in real-time. This research utilizes data mining technologies and new pattern matching algorithm. A clustering technique is used to get dynamic user segmentations using user preference to contents categories. Then a sequential pattern mining technique is used to extract contents access patterns in the user segmentations. And the recommendation is given by our recommendation algorithm using user contents preference history and contents access patterns of the segment. In the framework, preprocessing and data transformation and transition are implemented on DBMS. The proposed system is implemented to show that the framework is feasible. In the experiment using real-world large data, personalized recommendation is given in almost real-time and shows acceptable correctness.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2016.04a
• /
• pp.408-411
• /
• 2016

애플리케이션에서 고객들에 의해 생성된 평가정보는 해당 콘텐츠에 대한 고객별 선호도 정보로 볼 수 있기 때문에, 개인에게 맞춤형 추천 시스템을 설계하기 위해서 매우 중요하다. 현재 추천 시스템 분야에서 가장 많이 사용되고 있는 사용자 기반 추천 시스템은 사용자의 평점 정보만을 가지고 유사도를 측정하여 추천에 사용하고 있다. 그러나 이러한 평점 정보만을 가지고 사용자 유사도를 도출하는 것은 정밀하지 못할 수 있다. 따라서 본 연구에서는 사용자의 평점 정보 뿐만 아니라 콘텐츠의 내용을 활용하여 사용자의 선호 콘텐츠를 지식구조의 형태로 나타냄으로써 콘텐츠와 사용자의 관계를 유기적으로 표현하였다. 이와 같은 사용자의 지식구조를 바탕으로 사용자간의 유사도를 평가하고 추천에 활용하였고, 실험결과 제시된 방법으로 더 우수한 성능을 얻을 수 있는 것으로 나타났다.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2022.07a
• /
• pp.59-60
• /
• 2022

갈수록 개인화되어 가는 추천시스템은 다양한 모델에 의해 그 성능이 향상되고 있으며 최근 추세는 다른 분야와 마찬가지로 딥러닝 기반 모델을 적용하여 추천 품질을 향상하고 있다. 그러나 대다수의 추천시스템은 하나의 도메인에서 개별적으로 사용될 뿐, 유사도메인이나 상이한 도메인이나 모두 다른 도메인에서의 사용자 성향이나 아이템 유사성을 거의 또는 전혀 고려하지 않고 있다. 이는 추천결과의 sparsity와 cold-start 문제를 더 악화시키는 원인이 된다. 본 논문은 다양한 딥러닝 모델 적용 추천 모델 중 오토인코더 모델을 지역특화 협업에 적용한 모델을 간략하게 소개하고 해당 모델을 상이한 도메인 간의 적용하기 위한 첫 단계로 손실함수 부분에 대해 개념적으로 설명하고자 한다.

• The KIPS Transactions:PartD
• /
• v.16D no.2
• /
• pp.273-280
• /
• 2009

Today, many services and products that used to be only provided on offline have been being provided on the web according to the improvement of computing environment and the activation of web usage. These web-based services and products tend to be provided to customer by customer's preferences. This paradigm that considers customer's opinions and features in selecting is called personalization. The related research field is a recommendation. And this recommendation is performed by recommender system. Generally the recommendation is made from the preferences and tastes of customers. And recommender system provides this recommendation to user. However, the recommendation techniques have a couple of problems; they do not provide suitable recommendation to new users and also are limited to computing space that they generate recommendations which is dependent on ratings of products by users. Those problems has gathered some continuous interest from the recommendation field. In the case of new users, so similar users can't be classified because in the case of new users there is no rating created by new users. The problem of the limitation of the recommendation space is not easy to access because it is related to moneywise that the cost will be increasing rapidly when there is an addition to the dimension of recommendation. Therefore, I propose the solution of the recommendation problem of new user and the usage of item quality as weight to improve the accuracy of recommendation in this paper.

• The Journal of the Korea Contents Association
• /
• v.10 no.11
• /
• pp.10-19
• /
• 2010

The progress of computers and Web has given rise to a rapid increase of the quantity of the useful information, which is making the demand of recommender systems widely expanding. Like in other domains, a recommender system in a digital library is important, but there are only a few studies about the recommender system of research papers, Moreover none is there in korea to our knowledge. In the paper, we seek for a way to develop the NDSL recommender system of research papers based on the survey of related studies. We conclude that NDSL needs to modify the way to collect user's interests from explicit to implicit method, and to use user-based and memory-based collaborative filtering mixed with contents-based filtering(CF). We also suggest the method to mix two filterings and the use of personal ontology to improve user satisfaction.

• The Journal of Society for e-Business Studies
• /
• v.22 no.3
• /
• pp.75-85
• /
• 2017

Personalization recommendations have already proven in many areas of the e-commerce industry. For personalization recommendations, additional work such as reclassifying items is generally necessary, which requires personal information. In this study, we propose a recommendation technique that neither exploit personal information nor reclassify items. We focus on music recommendation and performed experiments with actual music listening data. Experimental analysis shows that the proposed method may result in meaningful recommendations albeit it exploits less amount of data. We analyze the appropriate number of items and present future considerations for contextual recommendation.