a. Propose a design for a BPN to help make health insurance premium decisions for an insurance company. Prospective clients for health insurance are to be classified as low risk or high-risk candidates, where high-risk clients are to be charged a higher premium or may even be refused insurance. Carefully, discuss the appropriate inputs and outputs to your BPN.
What data representation do you recommend and what type of activation function would you use?
What is the initial architecture of your network?
b. Indicate where the necessary training data would come from.
c. In light of advances in computer and genetics technology, discuss some of the ethical and legal problems that could arise.
d. Describe your training methodology. Include a discussion of the types of problems that can occur and recommend possible remedies in each case. When should training stop?
e. Describe several approaches to validating your BPN. Give an advantage and disadvantage for each approach.