• Proceedings of the Korea Information Processing Society Conference
• /
• 2017.04a
• /
• pp.439-441
• /
• 2017

Matrix factorization은 사용자의 아이템 선호도를 통해 아이템을 추천해주는 성공적인 기술 중 하나이다. 이 기법은 사용자-아이템의 선호도 행렬을 채우는 것을 목표로 한다. 이 목표를 달성하기 위해 사용자-아이템의 선호도 행렬을 사용자 행렬(user latent factor)와 아이템 행렬(item latent factor)로 분해하고, 각 행렬에 대해 추론하여 완성된 사용자-아이템의 선호도 행렬을 추론한다. 하지만 Matrix factorization은 아이템의 수가 많고, 아이템에 대한 사용자들의 선호도 데이터가 적을 때 성능이 제한된다. 또한 새로운 아이템이 추가되었을 때, 새로운 아이템에 대한 사용자들의 선호도 정보가 없기 때문에 새로운 아이템이 추천되지 않는다는 문제를 가진다. 이를 해결하기 위해 본 논문에서는 아이템에 대한 부가적인 정보인 아이템 간의 유사도 정보와 아이템의 시나리오 정보의 유사도를 모델링하여 기존의 전통적인 Matrix factorization에 추가하는 아이템 정보 기반 추천 시스템을 제안한다.

• Proceedings of the Korean Statistical Society Conference
• /
• 2003.10a
• /
• pp.239-244
• /
• 2003

본 연구에서는 확률모형의 모수로부터 얻어지는 여러형태의 함수간의 크기를 다중비교 하는 방법을 제안하고자 한다. 이 방법은 비교대상인 모수 함수간의 선호확률을 베이지안 방법으로 추정하고, 이들로부터 얻어지는 선호행렬을 이용한 새로운 다중비교법이다. 이러한 방법의 제안에 필요한 이론과 비교기준을 고안하였으며, 응용 예로, 제안된 방법을 s개의 독립인 지수분포 모수의 기하평균 크기비교에 적용하였다.

• Journal of the Korea Society of Computer and Information
• /
• v.19 no.5
• /
• pp.61-69
• /
• 2014

Collaborative filtering recommendation creates similar item subset or similar user subset based on user preference about items and predict user preference to particular item by using them. Thus, if preference matrix has low density, reliability of recommendation will be sharply decreased. To solve these problems we suggest Hybrid Preference Prediction Technique Using Weighting based Data Reliability. Preference prediction is carried out by creating similar item subset and similar user subset and predicting user preference by each subset and merging each predictive value by weighting point applying model condition. According to this technique, we can increase accuracy of user preference prediction and implement recommendation system which can provide highly reliable recommendation when density of preference matrix is low. Efficiency of this system is verified by Mean Absolute Error. Proposed technique shows average 21.7% improvement than Hao Ji's technique when preference matrix sparsity is more than 84% through experiment.

• The Korean Journal of Applied Statistics
• /
• v.14 no.2
• /
• pp.305-320
• /
• 2001

다차원선호분석(mutidimensional preference analysis)은 여러 상품들에 대한 개인(또는 그룹)의 선호도를 알아보기 위한 분석방법으로 결과는 보통 2차원 그림으로 제공된다. 본 연구에서는 의미있는 두 가지 최적척도 기준을 제안하고 이와 연관된 행 및 열표시자를 유도하고 있으며, 아울러 사전지식을 반영하기 위해 부분수량화를 다차원선호분석에 도입하는 방법을 제시한다. 또한 본 연구에서 제시한 다차원분석기법들을 실제 인터넷 검색엔진에 대한 선호도 자료에 적용한다.

• The Korean Journal of Applied Statistics
• /
• v.19 no.3
• /
• pp.447-460
• /
• 2006

This paper develops a multiple comparison method for finding an optimal ordering of K geometric means of exponential parameters. This is based on the paired comparison experimental arrangement whose results can naturally be represented by a completely oriented graph. Introducing posterior preference probabilities and stochastic transitivity conditions to the graph, we obtain a new graphical model that yields criteria for the optimal ordering in the multiple comparison. Necessary theories involved in the method and some computational aspects are provided. Some numerical examples are given to illustrate the efficiency of the suggested method.

• Journal of Advanced Navigation Technology
• /
• v.18 no.5
• /
• pp.468-475
• /
• 2014

The recommend system is getting more difficult real time recommend by lager preference data set, computing power and recommend algorithm. For this reason, recommend system is proceeding actively one's studies toward distribute processing method of large preference data set. This paper studied distribute processing method of large preference data set using hadoop distribute processing platform and mahout machine learning library. The recommend algorithm is used Co-occurrence Matrix similar to item Collaborative Filtering. The Co-occurrence Matrix can do distribute processing by many node of hadoop cluster, and it needs many computation scale but can reduce computation scale by distribute processing. This paper has simplified distribute processing of co-occurrence matrix by changes over from four stage to three stage. As a result, this paper can reduce mapreduce job and can generate recommend file. And it has a fast processing speed, and reduce map output data.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.11 no.3
• /
• pp.337-344
• /
• 2000

The wavelet analysis technique is applied in the spectral domain to efficiently represent the multi-scale features of the impedance matrices. In this scheme, the 2-D quadtree decomposition (applying the wavelet transform to only the part of the matrix) method often used in image processing area is applied for a sparse moment matrix. CG(Conjugate-Gradient) method is also applied for saving memory and computation time of wavelet transformed moment matrix. Numerical examples show that for rectangular cylinder case the non-zero elements of the transformed moment matrix grows only as O($N^{1.6}$).

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.14 no.3
• /
• pp.111-117
• /
• 2014

In this paper I propose a simple optimal solution seeking algorithm to a stable marriage problem. The proposed algorithm firstly constructs an $n{\times}n$ matrix of the sum of each gender's preference of the other gender $p_{ij}$. It then selects the minimum sum preference $_{min}p_{ij}$ in the constructed matrix and deletes its corresponding row i and column j. This process is repeated until $i=0{\cap}j=0$, after which the algorithm compares initially or last chosen $_{min}p_{ij}$ its alternatives to finally determine one that yields the maximum marginal increase in preference. When applied to 7 stable marriage problems, the proposed algorithm has improved on initial solutions of existing algorithms.

• The Korean Journal of Applied Statistics
• /
• v.23 no.5
• /
• pp.933-941
• /
• 2010

Biplot is a modern graphical methodology allowing for the projection of high-dimensional data to a low-dimensional subspace that is rich in information on variation in the data, correlation among variables as well as class separation. For the construction of biplots, we use a BiplotGUI package in a free statistical software R with increasing popularity. Moreover, using data from questionnaires given to Chonbuk National University freshmen in 2009, the relationship between career goals and career maturity are studied by applying the biplot method.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.16 no.2
• /
• pp.164-171
• /
• 2006

Recently Digital multimedia broadcasting (DMB) has been available as a commercial service. The users sometimes have difficulty in finding their preferred multimedia contents and need to spend a lot of searching time finding them. They are even very likely to miss their preferred contents while searching for them. In order to solve the problem, we need a method for recommendation users preferred only minimum information. We propose an algorithm and a system for recommending users' preferred contents using preference transition probability from user's usage history. The system includes four agents: a client manager agent, a monitoring agent, a learning agent, and a recommendation agent. The client manager agent interacts and coordinates with the other modules, the monitoring agent gathers usage data for analyzing the user's preference of the contents, the learning agent cleans the gathered usage data and modeling with state transition matrix over time, and the recommendation agent recommends the user's preferred contents by analyzing the cleaned usage data. In the recommendation agent, we developed the recommendation algorithm using a user's preference transition probability for the contents. The prototype of the proposed system is designed and implemented on the WIPI(Wireless Internet Platform for Interoperability). The experimental results show that the recommendation algorithm using a user's preference transition probability can provide better performances than a conventional method.