Evaluating Prediction Performance of Link Functions in Diagnosis Models for Alcohol Addiction

다국어 초록 (Multilingual Abstract)

Since diagnosis of alcohol addiction is largely questionnaire-based, the information communication technology (ICT)-driven advancement of big data application enables development of mathematical models for more effective diagnosis of addiction. This study presents a prototype diagnostic mathematical model for an interdisciplinary alcohol addiction diagnostic system under construction. Survey data acquired from 253 subjects using the Korean alcohol addiction test developed by the Korean National Mental Health Center was utilized to develop a mathematical model based on ordinal logistic analysis which defines the degree of alcohol addiction as probability. Because the type of link function determines the model’s accuracy, five types of link functions such as logit function, cauchit function, complementary log-log function, negative log-log function, and probit function were used to develop five types of diagnostic mathematical models. These models were then assessed for accuracy using prediction accuracy, probability of detection, and false alarm rate tests to select an optimal model for alcohol addiction diagnosis. Our study shows the resistance distribution of alcoholism is similar to the Gumbel distribution, and a model which uses the complementary log-log function is the most suitable one for the diagnosis of alcoholism.

참고문헌 (Reference)

1 손건태, "세 범주 예보모형에서 문턱값 결정을 위한 예측성 평가측도 선정 가이던스 개발" 한국자료분석학회 8 (8): 2553-2565, 2006

2 손건태, "봄철 천식에 대한 환자-맞춤형 이 범주 예측모형" 한국자료분석학회 13 (13): 2867-2875, 2011

3 이혜경, "보건진료소를 방문한 노인의 인지기능에 영향을 미치는 건강관련요인 : 식습관, 일상생활수행능력, 음주, 흡연을 중심으로" 한국자료분석학회 18 (18): 965-979, 2016

4 김용수, "로지스틱 회귀분석과 의사결정나무 분석을 이용한 데이터 요금제의 해지율 예측모형 수립 - 국내 모 이동통신업체의 사례연구 -" 한국자료분석학회 8 (8): 1915-1926, 2006

5 Cox, D. R., "The regression analysis of binary sequences" 20 : 215-242, 1958

6 Liu, I., "The analysis of ordered categorical data : An overview and a survey of recent developments" 14 (14): 1-73, 2005

7 McCabe, S. E., "Stressful events and other predictors of remission from drug dependence in the United States : Longitudinal results from a national survey" 71 : 41-47, 2016

8 McCullagh, P., "Regression models for ordinal data" 42 : 109-142, 1980

9 Mei, S., "Problematic Internet use, well-being, self-esteem and self-control : Data from a high-school survey in China" 61 : 74-79, 2016

10 Guisan, A., "Ordinal response regression models in ecology" 11 (11): 617-626, 2000

11 Armstrong, B. G., "Ordinal regression models for epidemiologic data" 129 (129): 191-204, 1989

12 Lesaffe, E., "Multiple-group logistic regression diagnostics" 38 : 425-440, 1989

13 Changpetch, P., "Model selection for logistic regression via association rules analysis" 83 (83): 1415-1428, 2013

14 Simonff, J. S., "Logistic regression, categorical predictors and goodness-of-fit: It depends on who you ask" 52 : 10-14, 1998

15 Barley, S. C., "Logistic regression in the medical literature : Standards for use and reporting, with particular attention to one medical domain" 54 (54): 979-985, 2001

16 Pregibon, D., "Logistic Regression diagnostics" 9 : 705-724, 1981

17 Katikireddi, S. V., "Has childhood smoking reduced following smoke-free public places legislation? A segmented regression analysis of cross-sectional UK school-based surveys" 18 (18): 1670-1674, 2016

18 Constantinou, A. C., "From complex questionnaire and interviewing data to intelligent Bayesian network models for medical decision support" 67 : 75-93, 2016

19 Hosmer, D. W., "A goodness-of-fit test for the multiple logistic regression model" 10 : 1043-1069, 1980

20 신양규, "A Bayesian Network Approach for Analyzing Causal Relationships of Questions in Alcoholism Questionnaire Survey" 한국자료분석학회 18 (18): 1207-1215, 2016