• Journal of Intelligence and Information Systems
• /
• v.16 no.4
• /
• pp.99-112
• /
• 2010

Ensemble learning is a method for improving the performance of classification and prediction algorithms. It is a method for finding a highly accurateclassifier on the training set by constructing and combining an ensemble of weak classifiers, each of which needs only to be moderately accurate on the training set. Ensemble learning has received considerable attention from machine learning and artificial intelligence fields because of its remarkable performance improvement and flexible integration with the traditional learning algorithms such as decision tree (DT), neural networks (NN), and SVM, etc. In those researches, all of DT ensemble studies have demonstrated impressive improvements in the generalization behavior of DT, while NN and SVM ensemble studies have not shown remarkable performance as shown in DT ensembles. Recently, several works have reported that the performance of ensemble can be degraded where multiple classifiers of an ensemble are highly correlated with, and thereby result in multicollinearity problem, which leads to performance degradation of the ensemble. They have also proposed the differentiated learning strategies to cope with performance degradation problem. Hansen and Salamon (1990) insisted that it is necessary and sufficient for the performance enhancement of an ensemble that the ensemble should contain diverse classifiers. Breiman (1996) explored that ensemble learning can increase the performance of unstable learning algorithms, but does not show remarkable performance improvement on stable learning algorithms. Unstable learning algorithms such as decision tree learners are sensitive to the change of the training data, and thus small changes in the training data can yield large changes in the generated classifiers. Therefore, ensemble with unstable learning algorithms can guarantee some diversity among the classifiers. To the contrary, stable learning algorithms such as NN and SVM generate similar classifiers in spite of small changes of the training data, and thus the correlation among the resulting classifiers is very high. This high correlation results in multicollinearity problem, which leads to performance degradation of the ensemble. Kim,s work (2009) showedthe performance comparison in bankruptcy prediction on Korea firms using tradition prediction algorithms such as NN, DT, and SVM. It reports that stable learning algorithms such as NN and SVM have higher predictability than the unstable DT. Meanwhile, with respect to their ensemble learning, DT ensemble shows the more improved performance than NN and SVM ensemble. Further analysis with variance inflation factor (VIF) analysis empirically proves that performance degradation of ensemble is due to multicollinearity problem. It also proposes that optimization of ensemble is needed to cope with such a problem. This paper proposes a hybrid system for coverage optimization of NN ensemble (CO-NN) in order to improve the performance of NN ensemble. Coverage optimization is a technique of choosing a sub-ensemble from an original ensemble to guarantee the diversity of classifiers in coverage optimization process. CO-NN uses GA which has been widely used for various optimization problems to deal with the coverage optimization problem. The GA chromosomes for the coverage optimization are encoded into binary strings, each bit of which indicates individual classifier. The fitness function is defined as maximization of error reduction and a constraint of variance inflation factor (VIF), which is one of the generally used methods to measure multicollinearity, is added to insure the diversity of classifiers by removing high correlation among the classifiers. We use Microsoft Excel and the GAs software package called Evolver. Experiments on company failure prediction have shown that CO-NN is effectively applied in the stable performance enhancement of NNensembles through the choice of classifiers by considering the correlations of the ensemble. The classifiers which have the potential multicollinearity problem are removed by the coverage optimization process of CO-NN and thereby CO-NN has shown higher performance than a single NN classifier and NN ensemble at 1% significance level, and DT ensemble at 5% significance level. However, there remain further research issues. First, decision optimization process to find optimal combination function should be considered in further research. Secondly, various learning strategies to deal with data noise should be introduced in more advanced further researches in the future.

• The Journal of Korean Society for Radiation Therapy
• /
• v.19 no.2
• /
• pp.99-106
• /
• 2007

Purpose: The patient's position and anatomy during the treatment course little bit varies to some extend due to setup uncertainties and organ motions. These factors could affected to not only the dose coverage of the gross tumor but over dosage of normal tissue. Setup uncertainties and organ motions can be minimized by precise patient positioning and rigid immobilization device but some anatomical site such as prostate, the internal organ motion due to physiological processes are challenge. In planning procedure, the clinical target volume is a little bit enlarged to create a planning target volume that accounts for setup uncertainties and organ motion as well. These uncertainties lead to differences between the calculated dose by treatment planning system and the actually delivered dose. The purpose of this study was to evaluate the differences of interfractional displacement of organ and GTV based on the tomoimages. Materials and Methods: Over the course of 3 months, 3 patients, those who has applied rectal balloon, treated for prostatic cancer patient's tomoimage were studied. During the treatment sessions 26 tomoimages per patient, Total 76 tomoimages were collected. Tomoimage had been taken everyday after initial setup with lead marker attached on the patient's skin center to comparing with C-T simulation images. Tomoimage was taken after rectal balloon inflated with 60 cc of air for prostate gland immobilization for daily treatment just before treatment and it was used routinely in each case. The intrarectal balloon was inserted to a depth of 6 cm from the anal verge. MVCT image was taken with 5 mm slice thickness after the intrarectal balloon in place and inflated. For this study, lead balls are used to guide the registration between the MVCT and CT simulation images. There are three image fusion methods in the tomotherapy, bone technique, bone/tissue technique, and full image technique. We used all this 3 methods to analysis the setup errors. Initially, image fusions were based on the visual alignment of lead ball, CT anatomy and CT simulation contours and then the radiation therapist registered the MVCT images with the CT simulation images based on the bone based, rectal balloon based and GTV based respectively and registered image was compared with each others. The average and standard deviation of each X, Y, Z and rotation from the initial planning center was calculated for each patient. The image fusions were based on the visual alignment of lead ball, CT anatomy and CT simulation contours. Results: There was a significant difference in the mean variations of the rectal balloon among the methods. Statistical results based on the bone fusion shows that maximum x-direction shift was 8 mm and 4.2 mm to the y-direction. It was statistically significant (P=<0.0001) in balloon based fusion, maximum X and Y shift was 6 mm, 16mm respectively. One patient's result was more than 16 mm shift and that was derived from the rectal expansions due to the bowl gas and stool. GTV based fusion results ranging from 2.7 to 6.6 mm to the x-direction and 4.3$\sim$7.8 mm to the y-direction respectively. We have checked rotational error in this study but there are no significant differences among fusion methods and the result was 0.37$\pm$0.36 in bone based fusion and 0.34$\pm$0.38 in GTV based fusion.

• The Korean Journal of Nuclear Medicine Technology
• /
• v.20 no.1
• /
• pp.37-41
• /
• 2016

Purpose The importance of quality control by the error to a minimum, which for the purpose of enhancing the reliability of the examination is not be emphasized excess. Currently, most nuclear medicine laboratory are conducting the internal and external quality control, and they are applying the Levey-Jennings or Westgard Multi-Rules by using the commercialized quality control materials. The reliability of the nuclear medicine blood test which affects the diagnosis of patients and the treatment policy is being secured through this quality control activity. Therefore, researchers will evaluate the utility of the statistic quality control using the population distribution of the nuclear medicine blood test conducted targeting the checkup examinees by the additional technique of the reliability improvement. Materials and Methods A statistic analysis was performed about 12 items of the nuclear medicine blood test targeting 41,341 peoples who used the health screening and promotion center in Asan Medical Center from January, 2014 to December, 2014. The results of 12 items of the nuclear medicine blood test was divided into the monthly percentage of three groups: within reference values, over reference, and under reference to analyze the average value of the population distribution, standard deviation, and standard deviation index (SDI). Results The standard deviation of the population distribution mostly showed a result within ${\pm}2SD$ in all groups. However, When the standard deviation of the population distribution represented a result over ${\pm}2SD$, it was confirmed SDI was showing a result of SDI > -2 or SDI > 2. As a result of analyzing the population distribution of 12 items(AFP, CEA, CA19-9, CA125, PSA, TSH, FT4, Anti-Tg-Ab, Anti-TPO-Ab, Calcitonin, 25-OH-VitD3, Insulin) of the nuclear medicine blood part basic test, when SDI of the monthly percentage which deviated from the reference values was over ${\pm}2.0$, CA19-9 September was 2.2, Anti-Tg-Ab may was 2.2, Insulin January was 2.3, Insulin March was 2.4. It was confirmed these cases were attributed to the abnormality of the test reagent (maximum combination rate of isotope reagent declined) and the decline of the test response time. Conclusion The population distribution includes the entire attribute which becomes the study object. It is expected the statistic quality management using the population distribution which was conducted targeting the checkup examinees by dividing into three groups: within reference values, over reference, and under reference by means of this characteristics will be able to play a role of complementing the internal quality control program which is being carried out in the laboratory.