Prediction of the Probability of Customer Attrition by Using Cox Regression

Kang, Hyuncheol,Han, Sang-Tae The Korean Statistical Society 2004 Communications for statistical applications and me Vol.11 No.2

This paper presents our work on constructing a model that is intended to predict the probability of attrition at specified points in time among customers of an insurance company. There are some difficulties in building a data-based model because a data set may contain possibly censored observations. In an effort to avoid such kind of problem, we performed logistic regression over specified time intervals while using explanatory variables to construct the proposed model. Then, we developed a Cox-type regression model for estimating the probability of attrition over a specified period of time using time-dependent explanatory variables subject to changes in value over the course of the observations.

A Clustering Algorithm Considering Structural Relationships of Web Contents

Kang Hyuncheol,Han Sang-Tae,Sun Young-Su 한국통계학회 2005 Communications for statistical applications and me Vol.12 No.1

Application of data mining techniques to the world wide web, referred to as web mining, has been the focus of several recent researches. With the explosive growth of information sources available on the world wide web, it has become increasingly necessary to track and analyze their usage patterns. In this study, we introduce a process of pre-processing and cluster analysis on web log data and suggest a distance measure considering the structural relationships between web contents. Also, we illustrate some real examples of cluster analysis for web log data and look into practical application of web usage mining for eCRM.

Interpretation of Data Mining Prediction Model Using Decision Tree

Kang, Hyuncheol,Han, Sang-Tae,Choi, Jong-Ho 한국통계학회 2000 Communications for statistical applications and me Vol.7 No.3

Data mining usually deal with undesigned massive data containing many variables for which their characteristics and association rules are unknown, therefore it is actually not easy to interpret the results of analysis. In this paper, it is shown that decision tree can be very useful in interpreting data mining prediction model using two real examples.

Higher-order solutions for generalized canonical correlation analysis

Kang, Hyuncheol The Korean Statistical Society 2019 Communications for statistical applications and me Vol.26 No.3

Generalized canonical correlation analysis (GCCA) extends the canonical correlation analysis (CCA) to the case of more than two sets of variables and there have been many studies on how two-set canonical solutions can be generalized. In this paper, we derive certain stationary equations which can lead the higher-order solutions of several GCCA methods and suggest a type of iterative procedure to obtain the canonical coefficients. In addition, with some numerical examples we present the methods for graphical display, which are useful to interpret the GCCA results obtained.

Minimax Eccentricity Estimation for Multiple Set Factor Analysis

Hyuncheol Kang,Keeyoung Kim 한국통계학회 2002 Journal of the Korean Statistical Society Vol.31 No.2

간호학 연구에서 효과크기의 사용에 대한 고찰

강현철(Kang, Hyuncheol),연규필(Yeon, Kyupil),한상태(Han, Sang-Tae) 한국간호과학회 2015 Journal of Korean Academy of Nursing Vol.45 No.5

Purpose: The purpose of this study was to introduce the main concepts of statistical testing and effect size and to provide researchers in nursing science with guidance on how to calculate the effect size for the statistical analysis methods mainly used in nursing. Methods: For t-test, analysis of variance, correlation analysis, regression analysis which are used frequently in nursing research, the generally accepted definitions of the effect size were explained. Results: Some formulae for calculating the effect size are described with several examples in nursing research. Furthermore, the authors present the required minimum sample size for each example utilizing G<SUP>*</SUP>Power 3 software that is the most widely used program for calculating sample size. Conclusion: It is noted that statistical significance testing and effect size measurement serve different purposes, and the reliance on only one side may be misleading. Some practical guidelines are recommended for combining statistical significance testing and effect size measure in order to make more balanced decisions in quantitative analyses.

유클리드거리에 기초한 대응분석에서 평균 제거와 균등화의 영향

강현철(Hyuncheol Kang) 한국자료분석학회 2018 Journal of the Korean Data Analysis Society Vol.20 No.1

대응분석은 범주형 변수들에 의한 분할표 자료의 행과 열을 저차원 공간상에 나타내어, 그들의 관계를 시각적으로 탐색할 수 있는 유용한 자료분석 기법이다. 또한 대응분석은 분할표 자료가 아닌 어떤 형식의 2차원 표에도 적용할 수 있다는 장점을 가지고 있다. 대응분석에 의한 행렬도는 많은 분야에서 자주 사용되지만, 대부분의 경우 카이제곱거리에 기초한 대응분석이 주로 사용되는 경향이 있으며, 카이제곱거리에 기초한 대응분석에서는 상대적인 거리가 표현된다는 것을 주목할 필요가 있다. 이 논문에서는 카이제곱거리 대신에 유클리드거리에 기초한 대응분석의 개념과 절차를 자세히 소개하였으며, 실제 자료에 대한 사례분석을 통해 여러 가지 표준화 및 균등화 방법이 분석결과에 미치는 영향을 설명하였다. 이러한 결과를 통해 유클리드거리를 이용한 대응분석이 실제 문제에 사용될 때의 유용성을 제시하고자 한다. Correspondence analysis is a useful data analysis technique that can visualize the relationship between the rows and columns of contingency table by categorical variables on the low dimensional space. It also has the advantage that the correspondence analysis can be applied to any type of two-dimensional table. It should be noted that although the biplot by the correspondence analysis is often used in many fields, in most cases the corresponding analysis based on the chi-square distance tends to be used predominantly and the relative distance is expressed in the corresponding analysis based on the chi-square distance. In this paper, the concept and procedure of the correspondence analysis based on Euclidean distance instead of the chi-square distance are introduced in detail, and illustrations of actual data show the effect of various standardization and equalization methods on the analysis results. These results show that the correspondence analysis based on Euclidean distance can be useful when used in actual problems and that the correspondence analysis can be applied to various data in many fields.

순서형 프로빗 모형과 깁스표본 기법을 이용한 고객만족의 측정 및 평가

강현철(Hyuncheol Kang) 한국자료분석학회 2004 Journal of the Korean Data Analysis Society Vol.6 No.1

마케팅 리서치에 있어서 관심 대상이 되는 반응변수가 순서형으로 측정되는 경우가 자주 있으며, 설명변수에 대한 결측값의 처리와 모수에 대한 중요한 사전정보를 동시에 고려해야 하는 경우가 많다. 본 연구에서는 이러한 문제를 해결하기 위한 방법 중 하나로서 순서형 프로빗 모형 및 깁스표본 기법의 주요 개념을 소개하였다. 또한 이러한 기법들을 보다 일반화하여 제시하고, 이를 국내 한 기업에서 고객만족의 평가를 위해 수집된 자료에 적용함으로써 마케팅 관련 연구자들에게 하나의 전형을 제시하고자 한다. In marketing research, the response variable is often not merely continuous or categorical but inherently ordered scale. Also, it is needed that the missing data can be accommodated and managerial knowledge and natural constraints on the model parameter is incorporate. In this paper, we provide a framework and approaches to effectively estimate customer satisfaction model by using the ordered probit model and Gibbs sampling techniques.

구성타당도 평가에 있어서 요인분석의 활용

강현철(Kang, Hyuncheol) 한국간호과학회 2013 Journal of Korean Academy of Nursing Vol.43 No.5

Purpose: The purpose of this study is to provide researchers with a simplified approach to undertaking exploratory factor analysis for the assessment of construct validity. Methods: All articles published in 2010, 2011, and 2012 in Journal of Korean Academy of Nursing were reviewed and other relevant books and articles were chosen for the review. Results: In this paper, the following were discussed: preliminary analysis process of exploratory factor analysis to examine the sample size, distribution of measured variables, correlation coefficient, and results of KMO measure and Bartlett"s test of sphericity. In addition, other areas to be considered in using factor analysis are discussed, including determination of the number of factors, the choice of rotation method or extraction method of the factor structure, and the interpretation of the factor loadings and explained variance. Conclusion: Content validity is the degree to which elements of an assessment instrument are relevant to and representative of the targeted construct for a particular assessment purpose. This measurement is difficult and challenging and takes a lot of time. Factor analysis is considered one of the strongest approaches to establishing construct validity and is the most commonly used method for establishing construct validity measured by an instrument.

Detection of Zombie PCs Based on Email Spam Analysis

( Hyuncheol Jeong ),( Huy Kang Kim ),( Sangjin Lee ),( Eunjin Kim ) 한국인터넷정보학회 2012 KSII Transactions on Internet and Information Syst Vol.6 No.5

While botnets are used for various malicious activities, it is well known that they are widely used for email spam. Though the spam filtering systems currently in use block IPs that send email spam, simply blocking the IPs of zombie PCs participating in a botnet is not enough to prevent the spamming activities of the botnet because these IPs can easily be changed or manipulated. This IP blocking is also insufficient to prevent crimes other than spamming, as the botnet can be simultaneously used for multiple purposes. For this reason, we propose a system that detects botnets and zombie PCs based on email spam analysis. This study introduces the concept of “group pollution level” - the degree to which a certain spam group is suspected of being a botnet - and “IP pollution level” - the degree to which a certain IP in the spam group is suspected of being a zombie PC. Such concepts are applied in our system that detects botnets and zombie PCs by grouping spam mails based on the URL links or attachments contained, and by assessing the pollution level of each group and each IP address. For empirical testing, we used email spam data collected in an "email spam trap system" - Korea`s national spam collection system. Our proposed system detected 203 botnets and 18,283 zombie PCs in a day and these zombie PCs sent about 70% of all the spam messages in our analysis. This shows the effectiveness of detecting zombie PCs by email spam analysis, and the possibility of a dramatic reduction in email spam by taking countermeasure against these botnets and zombie PCs.