Reference no: EM133359659

The purpose of the predictive models is to help make investment decisions on loans. How will you evaluate the models on this business objective? Consider a simplified scenario - for example, that you have $100 to invest in each loan, based on the model's prediction. So, you will invest in all loans that are predicted to be 'Fully Paid'. Key questions here are: how much, on average, can you expect to earn after 3 years from a loan that is paid off, and what is your potential loss from a loan that has to be charged off ?
One can consider the average interest rate on loans for expected profit - is this a good estimate of your profit from a loan? For example, suppose the average int_rate in the data is 11.2%; so after 3 years, the $100 will be worth (100 + 3*11.2) = 133.6, i.e a profit of $33.6. Now, is 11.2% a reasonable value to expect - consider, if loans are paid back early, then the same rate of return will not continue till the end of the loan term. The data sample includes the actualTerm and actualReturn variables - these can be used to determine an average profit value. Explain what value of profit you use and why.

For a loan that is charged off, will the loss be the entire invested amount of $100? The data shows that such loans have do show some partial returned amount. Looking at the returned amount (actualReturn) for charged off loans, what proportion of invested amount can you expect to recover? Explain what value of loss you use.

You should also consider the alternate option of investing in, say in bank CDs (certificate of deposit); let's assume that this provides an interest rate of 2%. Then, if you invest $100, you will receive $106 after 3 years (not considering reinvestments, etc), for a profit of $6.Considering a confusion matrix, we can then have profit/loss amounts with each cell, as follows:

(a) Explain how you determine the value of profit, and value of loss you use for this analyses.