Question 6: Regression (2+2+2+2+2marks) This question uses the car insurance example described on page 3. The output below comes from running a linear regression in the R statistical environment where the response (dependent) variable was claim and the predictors consisted of each of the other variables as well as the square of age. (A) Briefly summarise in words how each predictor is associated with the occurrence of claims. (B) What is the predicted value of the claims variable for a new customer with the same predictor as ID 2 ? (C) Suggest one advantage and one disadvantage of using linear regression rather than a decision tree to predict claims. How could you modify the regression model to reduce the disadvantage, while still using a linear model? (D) The claims variable is binary, taking on the values 0 and 1. Suggest an alternative type of regression that could be used to reflect this, and in one or two sentences explain why this would be preferable to linear regression here. (E) Comparing a decision tree or linear regression with SVM, what are their advantages and disadvantages? Precursor to question 5 and 6: An Insurance Example This example will be used in Questions 5 and 6. An insurance company wants to predict which car insurance customers are likely to make a large claim (defined as $20,000 or more) in the next 12 months. They will use this to charge some customers a higher premium. They have compiled a dataset of 5000 past customers, recording some attributes for each customer at the start of a 12 month period, and whether or not they made a large claim in the 12 months following this start date. The first few records of this dataset are shown below: The data dictionary for the dataset is