Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: THEend8_
FINAL EXAMINATION
QUESTION 1 10 MARKS
Suppose you work as an actuary and want to model the 5-year mortality rates (i.e., the probability of death within five years) for a group of 1,000 life insurance policyholders aged 50. You have the following information for each:
Gender
Type of work (manual or non-manual);
Living area (High affluence, Middle class, Deprived);
Health condition (Terminally ill, Sick, healthy)
The following classification tree is obtained for the binary variable denoting the Status of the policyholder as observed in the data. The Status variable takes a value equal to 0 if the policyholder is alive at the end of the study (5 years, where the policyholders are observed starting at the age of 50 until age 55) and 1 if dead.
Required:
With reference to the above case, please answer all of the following questions:
Part A [max 100 words] (5 marks)
What does each number in a node represent? Please describe the results in the node circled in red.
Part B [max 100 words] (5 marks)
What are the most influential variables characterising policyholders’ mortality in order of importance?
QUESTION 2 30 MARKS
A financial institution has decided to implement a Machine Learning (ML) algorithm to automate its loan approval process. The algorithm, developed by a third-party AI solutions provider, is designed to analyse a range of consumer data to determine loan eligibility and terms. To maintain a transparent process, the firm offers applicants the option to discuss the loan outcome decision with a representative. The representatives are well-trained in explaining how the applicant's data influenced the loan decision.
The third-party developer has provided the firm with two versions of the algorithm. The first version is "gender-blind," designed to exclude any variables that could directly or indirectly reveal an applicant's gender. The second version incorporates gender-specific features, which can enhance the accuracy of loan predictions based on statistical differences in financial behaviour across genders. The third-party developer is also open to feedback on variables to include in the model and how to determine who is loan-worthy.
Required:
With reference to the above case, please answer all of the following questions:
Part A [max 100 words] (4 marks)
Given a choice between the gender-blind version and the gender-specific version of the loan decision algorithm, identify which option the firm should choose to use and any general adjustments that could be made to make the algorithm more responsible.
Part B [max 200 words] (6 marks)
Considering the scenario, identify and discuss the issues that raise concerns under the Australian Responsible AI Principles, focusing specifically on fairness, accountability, transparency, and contestability.