Titanic-Survival-Prediction

The popular Titanic dataset ia available and as part of practice competition in Kaggle. For the full HTML page output, please click this link.

Data Exploratory

In the Rmd script, I will use library(Amelia) for quick visualization on the missing data, and another code to actual get the number of the missing data points. There are also some other library which I used (apart from ggplot2) to visualize the relationship of the feature to survival binary output.

Sample code and output

library(vcd)
mosaicplot(training.data$pclass ~ training.data$survived, 
           main="Passenger Fate by Traveling Class", shade=FALSE, 
           color=TRUE, xlab="Passenger class", ylab="Survived")

Prediction

Finally we will use randomForest library to make the prediction as well as party::cforest and compare the result.

Outcome from randomForest library

confusionMatrix(test.batch.rf.pred, test.batch$survived)

## Confusion Matrix and Statistics
## 
##           Reference
## Prediction  0  1
##          0 96 18
##          1 13 50
##                                         
##                Accuracy : 0.825         
##                  95% CI : (0.761, 0.878)
##     No Information Rate : 0.616         
##     P-Value [Acc > NIR] : 1.3e-09       
##                                         
##                   Kappa : 0.625         
##  Mcnemar's Test P-Value : 0.472         
##                                         
##             Sensitivity : 0.881         
##             Specificity : 0.735         
##          Pos Pred Value : 0.842         
##          Neg Pred Value : 0.794         
##              Prevalence : 0.616         
##          Detection Rate : 0.542         
##    Detection Prevalence : 0.644         
##       Balanced Accuracy : 0.808         
##                                         
##        'Positive' Class : 0             
## 

Outcome from party library (unbiased forest)

confusionMatrix(test.batch.party.pred, test.batch$survived)
## Confusion Matrix and Statistics
## 
##           Reference
## Prediction   0   1
##          0 102  16
##          1   7  52
##                                         
##                Accuracy : 0.87          
##                  95% CI : (0.811, 0.916)
##     No Information Rate : 0.616         
##     P-Value [Acc > NIR] : 6e-14         
##                                         
##                   Kappa : 0.718         
##  Mcnemar's Test P-Value : 0.0953        
##                                         
##             Sensitivity : 0.936         
##             Specificity : 0.765         
##          Pos Pred Value : 0.864         
##          Neg Pred Value : 0.881         
##              Prevalence : 0.616         
##          Detection Rate : 0.576         
##    Detection Prevalence : 0.667         
##       Balanced Accuracy : 0.850         
##                                         
##        'Positive' Class : 0             
##