• The Korean Journal of Applied Statistics
• /
• v.13 no.2
• /
• pp.343-352
• /
• 2000

In this study. the author proposes a nonhierarchical clustering method. called the "Double K-Means Clustering", which performs clustering of multivariate observations with the following algorithm: Step I: Carry out the ordinary K-means clmitering and obtain k temporary clusters with sizes $n_1$,... , $n_k$, centroids $c_$1,..., $c_k$ and pooled covariance matrix S. $\bullet$ Step II-I: Allocate the observation x, to the cluster F if it satisfies ..... where N is the total number of observations, for -i = 1, . ,N. $\bullet$ Step II-2: Update cluster sizes $n_1$,... , $n_k$, centroids $c_$1,..., $c_k$ and pooled covariance matrix S. $\bullet$ Step II-3: Repeat Steps II-I and II-2 until the change becomes negligible. The double K-means clustering is nearly "optimal" under the mixture of k multivariate normal distributions with the common covariance matrix. Also, it is nearly affine invariant, with the data-analytic implication that variable standardizations are not that required. The method is numerically demonstrated on Fisher's iris data.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.41 no.2
• /
• 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.

• The Journal of Society for e-Business Studies
• /
• v.10 no.2
• /
• pp.109-132
• /
• 2005

With the flood of multimedia contents over the digital TV channels, the internet, and etc., users sometimes have a difficulty in finding their preferred contents, spend heavy surfing time to find them, and are even very likely to miss them while searching. In this paper we suggests two-step clustering technique using time schema on how the system can recommend the user's preferred contents based on the collaborative filtering that has been proved to be successful when new users appeared. This method maps and recommends users' profile according to the gender and age at the first step, and then recommends a probabilistic item clustering customers who choose the same item at the same time based on time schema at the second stage. In addition, this has improved the accuracy of predictions in recommendation and the efficiency in time calculation by reflecting feedbacks of the result of the recommender engine and dynamically update customers' preference.

• Journal of KIISE:Software and Applications
• /
• v.26 no.12
• /
• pp.1468-1481
• /
• 1999

didates.본 논문은 절차 중심 소프트웨어를 객체 지향 소프트웨어로 재/역공학하기 위한 다단계 절차중 첫 절차인 객체 추출 절차에 대하여 기술한다. 사용한 객체 추출 방법은 전처리, 기본 분할 및 결합, 정제 결합, 결정 및 통합의 다섯 단계로 이루어진다 : 1) 전처리 과정에서는 객체 추출을 위한 FTV(Function, Type, Variable) 그래프를 생성/분할 및 클러스터링하고, 2) 기본 분할 및 결합 단계에서는 다중 객체 추출을 위한 그래프를 생성하고 생성된 그래프의 정적 객체를 추출하며, 3) 정제 결합 단계에서는 동적 객체를 추출하며, 4) 결정 단계에서는 영역 모델링과 다중 객체 후보군과의 유사도를 측정하여 영역 전문가가 하나의 최적합 후보를 선택할 수 있는 측정 결과를 제시하며, 5) 통합 단계에서는 전처리 과정에서 분리된 그래프가 여러 개 존재할 경우 각각의 처리된 그래프를 통합한다. 본 논문에서는 클러스터링 순서가 고정된 결정론적 방법을 사용하였으며, 가능한 경우의 수에 따른 다중 객체 후보, 객관적이고 의미가 있는 객체 추출 방법으로의 정제와 결정, 영역 모델링을 통한 의미적 관점에 기초한 방법 등을 사용한다. 이러한 방법을 사용함으로써 전문가는 객체 추출 단계에서 좀더 다양하고 객관적인 선택을 할 수 있다.Abstract This paper presents an object extraction process, which is the first phase of a methodology to transform procedural software to object-oriented software. The process consists of five steps: the preliminary, basic clustering & inclusion, refinement, decision and integration. In the preliminary step, FTV(Function, Type, Variable) graph for object extraction is created, divided and clustered. In the clustering & inclusion step, multiple graphs for static object candidate groups are generated. In the refinement step, each graph is refined to determine dynamic object candidate groups. In the decision step, the best candidate group is determined based on the highest similarity to class group modeled from domain engineering. In the final step, the best group is integrated with the domain model. The paper presents a new clustering method based on static clustering steps, possible object candidate grouping cases based on abstraction concept, a new refinement algorithm, a similarity algorithm for multiple n object and m classes, etc. This process provides reengineering experts an comprehensive and integrated environment to select the best or optimal object candidates.

• Korean Journal of Cognitive Science
• /
• v.11 no.2
• /
• pp.53-58
• /
• 2000

This paper reconstructs local region of a facial expression image from extracted feature points of facial expression image using FCM(Fuzzy C-Meang) clustering algorithm with Gabor wavelets. The feature extraction in a face is two steps. In the first step, we accomplish the edge extraction of main components of face using average value of 2-D Gabor wavelets coefficient histogram of image and in the next step, extract final feature points from the extracted edge information using FCM clustering algorithm. This study presents that the principal components of facial expression images can be reconstructed with only a few feature points extracted from FCM clustering algorithm. It can also be applied to objects recognition as well as facial expressions recognition.

• KEPCO Journal on Electric Power and Energy
• /
• v.2 no.1
• /
• pp.61-69
• /
• 2016

Understanding load patterns and customer classification is a basic step in analyzing the behavior of electricity consumers. To achieve that, there have been many researches about clustering customers' daily load data. Nowadays, the deployment of advanced metering infrastructure (AMI) and big-data technologies make it easier to study customers' load data. In this paper, we study load clustering from the view point of yearly and daily load pattern. We compare four clustering methods; K-means clustering, hierarchical clustering (average & Ward's method) and DBSCAN (Density-Based Spatial Clustering of Applications with Noise). We also discuss the relationship between clustering results and Korean Standard Industrial Classification that is one of possible labels for customers' load data. We find that hierarchical clustering with Ward's method is suitable for clustering load data and KSIC can be well characterized by daily load pattern, but not quite well by yearly load pattern.

• Proceedings of the Korean Society for Emotion and Sensibility Conference
• /
• 2000.04a
• /
• pp.126-132
• /
• 2000

This paper presents a facial expression recognition based on Gabor wavelets that uses a fuzzy C-means(FCM) clustering algorithm and neural network. Features of facial expressions are extracted to two steps. In the first step, Gabor wavelet representation can provide edges extraction of major face components using the average value of the image's 2-D Gabor wavelet coefficient histogram. In the next step, we extract sparse features of facial expressions from the extracted edge information using FCM clustering algorithm. The result of facial expression recognition is compared with dimensional values of internal stated derived from semantic ratings of words related to emotion. The dimensional model can recognize not only six facial expressions related to Ekman's basic emotions, but also expressions of various internal states.

• IE interfaces
• /
• v.11 no.1
• /
• pp.85-95
• /
• 1998

This study is focussed on an optimal vehicle routing model for multi-supply centers in two-echelon logistic system. The aim of this study is to deliver goods for demand sites with optimal decision. This study investigated an integrated model using step-by-step approach based on relationship that exists between the inventory allocation and vehicle routing with restricted amount of inventory and transportations such as the capability of supply centers, vehicle capacity and transportation parameters. Three sub-models are developed: 1) sector-clustering model, 2) a vehicle-routing model based on clustering and a heuristic algorithm, and 3) a vehicle route scheduling model using TSP-solver based on genetic and branch-and-bound algorithm. Also, we have developed computer programs for each sub-models and user interface with visualization for major inputs and outputs. The application and superior performance of the proposed model are demonstrated by several sample runs for the inventory-allocation and vehicle routing problems.

• The Journal of Society for e-Business Studies
• /
• v.26 no.4
• /
• pp.133-154
• /
• 2021

Today smart phone is the most common media with a vehicle called 'application'. In order to understand how media users select applications and build their repertoire, this study conducted two-step approach using big data from smart phone log for 4 weeks in November 2019, and finally classified 8 media repertoire groups. Each of the eight media repertoire groups showed differences in time spent of mobile application category compared to other groups, and also showed differences between groups in demographic distribution. In addition to the academic contribution of identifying the mobile application repertoire with large scale behavioral data, this study also has significance in proposing a two-step approach that overcomes 'outlier issue' in behavioral data by extracting prototype vectors using SOM (Sefl-Organized Map) and applying it to k-means clustering for optimization of the classification. The study is also meaningful in that it categorizes customers using e-commerce services, identifies customer structure based on behavioral data, and provides practical guides to e-commerce communities that execute appropriate services or marketing decisions for each customer group.

• Journal of the Korea Society of Computer and Information
• /
• v.13 no.6
• /
• pp.147-154
• /
• 2008

First, this paper compute difference value between frames using the composed method of $X^2$ histogram and color histogram and the normalization. Next, cluster representative frame was decided by using the clustering for distance and the k-mean grouping. Finally, representative frame of group was decided by using the likelihood ratio. Proposed method can be known by experiment as outstanding of detection rather than other methods, due to computing of difference value, clustering and grouping, and detecting of representative frame.