The predictions are in log form. Take the exponent and multiply by the base hazard estimate to obtain the predicted value at 6 months:
#head(pred_validation)# pred.val <- base[6,1]*exp(pred_validation)
Let's assume we want to predict the risk of churn that occurs halfway through the analysis period (time=6):
First, we will first take the exponents of the prediction and multiply them by the base hazard estimate that was shown in the hazard column at time=6. This method, in effect, adds the predictive power of the coefficients to the base hazard rate:
pred.val <- base[6,1]*exp(pred_validation)
We will now merge the predicted values in with the raw values from the test dataset and view the results after verifying ...