Comparisons of predictive modeling techniques for breast cancer in Korean women
2
조회
0
다운로드
3
인용
0
참고문헌
키워드
초록
Objective: To develop breast cancer prediction models and to compare their predictive performance by using Bayesian Networks (BN), Nave Bayes (NB), Classification and Regression Trees (CART), and Logistic Regression (LR). Methods: The dataset consisting of 109 breast cancer patients and 100 healthy women was used. Hugin ResearcherTM 6.7 and PoulinHugin 1.5, both of which are NB modeling software, were used. For the LR model and CART, ECMiner was used. Results: The highest area under the receiver operating characteristic curve (AUC) was shown in the Tree augmented NB model as .90. The lowest AUC was CART with .48; that of the LR model was .86. Two BN models with prior knowledge and without prior knowledge did not show any difference at all (.64 vs. .65). The lifts of four models (Simple NB, Tree Augmented NB, Hierarchical NB, LR) were 1.9. The AUCs in both the NB and LR models were higher than that of the previously established models that have been published by using LR methods. Conclusion: NB could be preferred to LR in the development of a predictive model to promote regular screening tests and early detection, which is more or less free from statistical assumptions and limitations.
Abstract
Objective: To develop breast cancer prediction models and to compare their predictive performance by using Bayesian Networks (BN), Nave Bayes (NB), Classification and Regression Trees (CART), and Logistic Regression (LR). Methods: The dataset consisting of 109 breast cancer patients and 100 healthy women was used. Hugin ResearcherTM 6.7 and PoulinHugin 1.5, both of which are NB modeling software, were used. For the LR model and CART, ECMiner was used. Results: The highest area under the receiver operating characteristic curve (AUC) was shown in the Tree augmented NB model as .90. The lowest AUC was CART with .48; that of the LR model was .86. Two BN models with prior knowledge and without prior knowledge did not show any difference at all (.64 vs. .65). The lifts of four models (Simple NB, Tree Augmented NB, Hierarchical NB, LR) were 1.9. The AUCs in both the NB and LR models were higher than that of the previously established models that have been published by using LR methods. Conclusion: NB could be preferred to LR in the development of a predictive model to promote regular screening tests and early detection, which is more or less free from statistical assumptions and limitations.
참고문헌 (31)
- Stewart BW, "World cancer reoprt" (2003), IARC Press
- "2002 Annual report of the Korea central registry" (2002)[URL]
- "1998 Annual report of the Korea central registry" (1998)[URL]
- The cause of death statistics, "KNSO", Korean National Statistical Office, ~ (2006)
- Annual report of the central cancer registry in Korea, "NCC", National Cancer Center, ~ (2002)
- The cause of death statistics, "KNSO", Korean National Statistical Office, ~ (2003)
- The status of health screening in 2003, "NHIC", National Health Insurance Corporation, ~ (2003)
- Park SK, "The estimation of breast cancer disease‐probability by difference of individual susceptibility", Cancer Research and Treatment, 35(1), pp.35~51 (2003)
- Lee EO, "Determining the main risk factors and high‐risk groups of breast cancer using a predictive model for breast cancer risk assessment in South Korea", Cancer Nursing, 27(5), pp.400~406 (2004)
- Lee SM, "Bayesian Network Approaches to Study Health Outcomes", Ph.D., pp.2003
- Arena VC, "The utility of structure‐activity relationship (SAR) models for prediction and covariate selection in developmental toxicity: comparative analysis of logistic regression and decision tree models 8", Sar & Qsar in Environmental Research, 15(1), pp.1~18 (2004)
- Sims CJ, "Predicting cesarean delivery with decision tree models 16", American Journal of Obstetrics & Gynecology, 183(5), pp.1198~1206 (2000)
- Pradhan M, "Knowledge engineering for large belief networks", Proceeding of the 1994 conference on uncertainty in AI, pp.484~490 (1994)
- Hamerly G, "Bayesian approaches to failure prediction for disk drives", Proceedings of the eighteenth international conference on machine learning, pp.202~209 (2001)
- Cao J, "A naive bayes model to predict coupling between seven transmembrane domain receptors and G‐proteins", Bioinformatics, 2, pp.234~240 (2003)
- Jiang L, "Lecture notes in computer science: Learning tree augmented naive bayes for ranking" (2005), Springer
- Langseth H, "Classification using hierarchical naive bayes models", Machine Learning Journal, 63(2) (2006)
- Zhang NL, "Latent variable discovery in classification models", Artificial Intelligence in Medicine, 30(3), pp.283~299 (2003)
- Menard S, "Applied logistic regression analysis. 2 ed" (2002), Sage Publications
- Tu JV, "Advantages and disadvantages of using artificial neural networks versus logistic regression for predicting medical outcomes", Journal of Clinical Epidemiology, 49, pp.1225~1232 (1996)
- De Ferrari L, "Mining housekeeping genes with a Naive Bayes classifier", BMC Genomics, 7, pp.277~ (2006)
- Lee S‐M, "Bayesian network approaches to health services research", The Journal of Korean Society in Medical Informatic, 12(1), pp.71~81 (2005)
- Tamada Y, "Utilizing evolutionary information and gene expression data for estimating gene networks with bayesian network models", J Bioinform Comput Biol, 3(6), pp.1295~1313 (2005)
- Maglogiannis I, "Risk analysis of a patient monitoring system using Bayesian Network modeling", J Biomed Inform, 39(6), pp.637~647 (2006)
- Sakai S, "Accuracy in the iagnostic prediction of acute appendicitis based on the bayesian network model", Methods Inf Med, 46(6), pp.723~726 (2007)
- Leegon J, "Predicting hospital admission for Emergency Department patients using a Bayesian network", AMIA Annu Symp Proc, pp.1022~ (2005)
- Sanders DL, "Prospective evaluation of a Bayesian Network for detecting asthma exacerbations in a Pediatric Emergency Department", AMIA Annu Symp Proc, pp.1085~ (2006)
- Sanders DL, "Detecting asthma exacerbations in a pediatric emergency department using a Bayesian network", AMIA Annu Symp Proc, pp.684~688 (2006)
- Michalowski W, "Using a Bayesian belief network model to categorize length of stay for radical prostatectomy patients", Health Care Manag Sci, 9(4), pp.341~348 (2006)