• Journal of the Korean Data and Information Science Society
• /
• 제22권4호
• /
• pp.741-753
• /
• 2011

마이크로어레이 유전자 발현데이터인 효모데이터를 이용하여 군집분석을 실시하였다. 모형기반 군집방법, K-평균법, 중앙값 중심분포 (PAM), 자기 조직화 지도 (SOM), 계층적 Ward 군집방법을 이용하여 군집화를 실시하고, 연결성 측도 (connectivity), Dunn지수, 실루엣 측도 (silhouette)를 이용하여 각 군집방법에 대한 유효성을 측정하고 군집분석 결과를 비교하고자한다.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• 제4권3호
• /
• pp.337-340
• /
• 2004

The clustering problem can be formulated as the problem to find the number of clusters and a partition matrix from a given data set using the iterative or non-iterative algorithms. The author proposes a nearest neighbor and validity-based clustering algorithm where each data point in the data set is linked with the nearest neighbor data point to form initial clusters and then a cluster in the initial clusters is linked with the nearest neighbor cluster to form a new cluster. The linking between clusters is continued until no more linking is possible. An optimal set of clusters is identified by using the conventional cluster validity index. Experimental results on well-known data sets are provided to show the effectiveness of the proposed clustering algorithm.

• 한국멀티미디어학회논문지
• /
• 제25권6호
• /
• pp.886-893
• /
• 2022

Outlier detection refers to the task of detecting data that deviate significantly from the normal data distribution. Most outlier detection methods compute an outlier score which indicates the degree to which a data sample deviates from normal. However, setting a threshold for an outlier score to determine if a data sample is outlier or normal is not trivial. In this paper, we propose a binary prediction method for outlier detection based on spectral clustering and one-class SVM ensemble. Given training data consisting of normal data samples, a clustering method is performed to find clusters in the training data, and the ensemble of one-class SVM models trained on each cluster finds the boundaries of the normal data. We show how to obtain a threshold for transforming outlier scores computed from the ensemble of one-class SVM models into binary predictive values. Experimental results with high dimensional text data show that the proposed method can be effectively applied to high dimensional data, especially when the normal training data consists of different shapes and densities of clusters.

• 산업경영시스템학회지
• /
• 제41권2호
• /
• pp.1-8
• /
• 2018

K-means algorithm is one of the most popular and widely used clustering method because it is easy to implement and very efficient. However, this method has the limitation to be used with fixed number of clusters because of only considering the intra-cluster distance to evaluate the data clustering solutions. Silhouette is useful and stable valid index to decide the data clustering solution with number of clusters to consider the intra and inter cluster distance for unsupervised data. However, this valid index has high computational burden because of considering quality measure for each data object. The objective of this paper is to propose the fast and simple speed-up method to overcome this limitation to use silhouette for the effective large-scale data clustering. In the first step, the proposed method calculates and saves the distance for each data once. In the second step, this distance matrix is used to calculate the relative distance rate ($V_j$) of each data j and this rate is used to choose the suitable number of clusters without much computation time. In the third step, the proposed efficient heuristic algorithm (Group search optimization, GSO, in this paper) can search the global optimum with saving computational capacity with good initial solutions using $V_j$ probabilistically for the data clustering. The performance of our proposed method is validated to save significantly computation time against the original silhouette only using Ruspini, Iris, Wine and Breast cancer in UCI machine learning repository datasets by experiment and analysis. Especially, the performance of our proposed method is much better than previous method for the larger size of data.

• 산업경영시스템학회지
• /
• 제43권2호
• /
• pp.25-32
• /
• 2020

Data on patent and scientific paper is considered as a useful information source for analyzing technological information and has been widely utilized. Technology big data is analyzed in various ways to identify the latest technological trends and predict future promising technologies. Clustering is one of the ways to discover new features by creating groups from technology big data. Patent includes refined bibliographic information such as patent classification code whereas scientific paper does not have appropriate bibliographic information for clustering. This research proposes a new approach for clustering data of scientific paper by utilizing reference titles in each scientific paper. In this approach, the reference titles are considered as textual information because each reference consists of the title of the paper that represents the core content of the paper. We collected the scientific paper data, extracted the title of the reference, and conducted clustering by measuring the text-based similarity. The results from the proposed approach are compared with the results using existing methodologies that one is the approach utilizing textual information from titles and abstracts and the other one is a citation-based approach. The suggested approach in this paper shows statistically significant difference compared to the existing approaches and it shows better clustering performance. The proposed approach will be considered as a useful method for clustering scientific papers.

• 응용통계연구
• /
• 제35권2호
• /
• pp.265-283
• /
• 2022

사용자들의 영화정보를 기록한 MovieLens 데이터는 추천 시스템 연구에서 아이디어를 탐색하고 검증하는데 상당한 가치가 있는 데이터로, 기존 데이터 분할 및 군집화 알고리즘을 사용하여 사용자 평점 데이터를 기반으로 항목 집합을 분할하는 연구 등에 사용되는 데이터이다. 본 논문에서는 기존 연구에서 대표적으로 사용되었던 영화 평점 데이터와 영화 장르 데이터를 통해 사용자의 장르 선호도를 예측하여 선호도 패턴을 기반으로 사용자를 군집화(clustering)하고, 유의미한 정보를 얻는 연구를 진행하였다. MovieLens 데이터는 영화의 전체 개수에 비해 사용자별 평균 영화 평점 수가 낮아 결측 비율이 높다. 이러한 이유로 기존의 군집화 방법을 적용하는 데 한계가 존재한다. 본 논문에서는 MovieLens 데이터 특성에 모티브를 얻어 쌍별 규합 벌점함수(pairwise fused penalty)를 활용한 볼록 군집화(convex clustering) 기반의 방법을 제안한다. 특히 결측치 대체(missing imputation)도 동시에 해결하는 최적화 문제를 통해 기존의 군집화 분석과 차별화하였다. 군집화는 반복 알고리즘인 ADMM을 통해 제안하는 최적화 문제를 풀어 진행한다. 또한 시뮬레이션과 MovieLens 데이터 적용을 통해 제안하는 군집화 방법이 기존의 방법보다 노이즈 및 이상치에 상대적으로 민감하지 않은 것으로 보인다.

• 한국지능시스템학회논문지
• /
• 제11권7호
• /
• pp.579-583
• /
• 2001

In the area of data clustering in high dimensional space, one of the difficulties is the time-consuming process for computing vector similarities. It becomes worse in the case of the agglomerative algorithm with the group-average link and mean centroid method, because the cluster similarity must be recomputed whenever the cluster center moves after the merging step. As a solution of this problem, we present an incremental method of similarity computation, which substitutes the scalar calculation for the time-consuming calculation of vector similarity with several measures such as the squared distance, inner product, cosine, and minimum variance. Experimental results show that it makes clustering speed significantly fast for very high dimensional data.

• 한국데이터정보과학회:학술대회논문집
• /
• 한국데이터정보과학회 2004년도 춘계학술대회
• /
• pp.51-60
• /
• 2004

K-means clustering is an iterative algorithm in which items are moved among sets of clusters until the desired set is reached. K-means clustering has been widely used in many applications, such as market research, pattern analysis or recognition, image processing, etc. It can identify dense and sparse regions among data attributes or object attributes. But k-means algorithm requires many hours to get k clusters that we want, because it is more primitive, explorative. In this paper we propose a new method of k-means clustering using a center of gravity for grid-based sample. It is more fast than any traditional clustering method and maintains its accuracy.

• Journal of Computing Science and Engineering
• /
• 제4권2호
• /
• pp.97-109
• /
• 2010

Nonnegative matrix factorization (NMF) is a popular method for multivariate analysis of nonnegative data, which is to decompose a data matrix into a product of two factor matrices with all entries restricted to be nonnegative. NMF was shown to be useful in a task of clustering (especially document clustering), but in some cases NMF produces the results inappropriate to the clustering problems. In this paper, we present an algorithm for orthogonal nonnegative matrix factorization, where an orthogonality constraint is imposed on the nonnegative decomposition of a term-document matrix. The result of orthogonal NMF can be clearly interpreted for the clustering problems, and also the performance of clustering is usually better than that of the NMF. We develop multiplicative updates directly from true gradient on Stiefel manifold, whereas existing algorithms consider additive orthogonality constraints. Experiments on several different document data sets show our orthogonal NMF algorithms perform better in a task of clustering, compared to the standard NMF and an existing orthogonal NMF.

• 한국지능시스템학회논문지
• /
• 제17권3호
• /
• pp.397-402
• /
• 2007

In this paper, we present a method for clustering incomplete Microarray data using alternating optimization in which a prior imputation method is not required. To reduce the influence of imputation in preprocessing, we take an alternative optimization approach to find better estimates during iterative clustering process. This method improves the estimates of missing values by exploiting the cluster Information such as cluster centroids and all available non-missing values in each iteration. The clustering results of the proposed method are more significantly relevant to the biological gene annotations than those of other methods, indicating its effectiveness and potential for clustering incomplete gene expression data.