Predictions

Now you are going to predict the survival curve for a new customer from the Cox Proportional Hazard model you estimated before. The model is still available in the object fitCPH.

The new customer is a woman and used a voucher in her first order (voucher = 1). The order was placed 21 days ago and had a shopping cart value of 99.90 dollars. She didn't return the order (returned = 0).

Remember: voucher and returned can have the values 0 or 1.

This exercise is part of the course

Machine Learning for Marketing Analytics in R

View Course

Exercise instructions

Create a one-row dataframe called newCustomer with the new customer's characteristics listed in the assignment text above.
Predict the expected median time until the second order for this customer using print() and plot the predicted survival curve.
You are informed that due to database problems the gender was incorrectly coded: The new customer is actually a man. The dataframe newCustomer is copied into a dataframe called newCustomer2. Now go ahead and change the respective variable to male.
Recompute the predicted median with the corrected data newCustomer2. What changed?

Hands-on interactive exercise

Have a go at this exercise by completing this sample code.

# Create data with new customer
___ <- data.frame(daysSinceFirstPurch = __, shoppingCartValue = ___, gender = "female", voucher = _, returned = _)

# Make predictions
pred <- survfit(fitCPH, newdata = ___)
print(___)
___(pred)

# Dataset is copied. Now correct the customer's gender there
newCustomer2 <- newCustomer
___$gender <- ___

# Redo prediction
pred2 <- ___(fitCPH, newdata = ___)
print(___)

Edit and Run Code