REVIEW OF NEURAL NETWORK METHODS IN SURVIVAL ANALYSIS

dc.contributor.authorSIN, KHENG SEANG
dc.date.accessioned2016-01-12T03:44:57Z
dc.date.available2016-01-12T03:44:57Z
dc.date.issued2005-06
dc.description.abstractIn recent years, neural networks have been applied in sur\'i\al analysis especially in medicine for the prediction of outcome variables such as survi\al times. The purpose of this study is to review neural networks which have been used in the survival analysis and compare the method with the Cox regression. There are many modified neural networks which have been used in sUf\'ival analysis to predict the survival curve or sUf\/ival times. In this study, we will use a similar approach to Brov.. n el al. (1997) to predict the survival curve. We will also use the sUf\!ival curve which is generated by Statistical Software-SPSS-Cox Regression to compare with the survival curve generated from the neural network. The weights which are generated from the neural network will be substituted into a program in MA TLAB to generate the sUf\/ival curve. We find that the method of Brown el al. (1997) can be used in survival analysis to predict the survival curve. We also used two separate multilayer perceptron neural networks, that is, one for censored data and another one for non-censored data to predict survival times without any modification. The purpose is to reduce the bias introduced by the censored data. The results showed that the estimated survival times are not very close to the true value. Therefore, some modification to the neural network models may be needed to predict the survival times. But, this method needs more time to be studied and can be considered as future research.en_US
dc.identifier.urihttp://hdl.handle.net/123456789/1452
dc.subjectREVIEW OF NEURALen_US
dc.subjectNETWORK METHODSen_US
dc.titleREVIEW OF NEURAL NETWORK METHODS IN SURVIVAL ANALYSISen_US
dc.typeThesisen_US
Files
License bundle
Now showing 1 - 1 of 1
Loading...
Thumbnail Image
Name:
license.txt
Size:
1.71 KB
Format:
Item-specific license agreed upon to submission
Description: