SURVIVAL MODEL AND ESTIMATION FOR LUNG CANCER PATIENTS - Quantitative Medical Data Analysis Using Mathematical Tools and Statistical Techniques - page 215

Fig. 2. Fit for Cumulative Hazard.

Fig. 3. Baseline Hazard as a Function of Time.

the output values assume only two possible values, we use logsig as the
nonlinear transfer function between layers.
When having traingda/learngdm as the training/learning function, the
NN reaches best performance, and the error rate for training set is 9%. The
error rate is defined as the rate of false “alive-dead” judgment for all 66
training cases. The network performance is shown in Figure 4.
After 6ma ixto

simulate 1000 patients' record. Each column of the matrix corre-
sponds to a covariate, and each row stores a patient's information
on PT,PN,STAGE,S − INT,D − INT ,and GRADE .Thenweusethe
trained NN to judge the STATUS of the patient, as we “believe” the NN

the NN is set up, we generate a 1000 ×