Exercise

# Precision, ROI, and AUC

The return on investment (ROI) can be decomposed into the precision multiplied by a ratio of return to cost. As discussed, it is possible for the precision of a model to be low, even while AUC of the ROC curve is high. If the precision is low, then the ROI will also be low. In this exercise, you will use a MLP to compute a sample ROI assuming a fixed `r`

, the return on a click per number of impressions, and `cost`

, the cost per number of impressions, along with precision and AUC of ROC curve values to check how the three values vary.

`X_train`

, `y_train`

, `X_test`

, `y_test`

are available in your workspace, along with `clf`

as a MLP classifier, probability scores stored in `y_score`

and predicted targets in `y_pred`

. `pandas`

as `pd`

and `sklearn`

are also available in your workspace.

Instructions

**100 XP**

- Calculate the precision
`prec`

of the MLP classifier. - Calculate the total ROI based on the precision
`prec`

.