QUALITY OF PREDICTION An employment firm has developed an analytic model that us

Question

QUALITY OF PREDICTION An employment firm has developed an analytic model that uses 25 characteristics of each job candidate (years of experience, highest degree, etc.) and 15 characteristics of each position (salary, equity, size of company, etc.) to predict whether a candidate will be successful in a position. "Success" is defined as a candidate holding the same or higher position within the company two years after joining. A placement was considered a "Failure" if the candidate was not with the company two years after joining or if the candidate was with the company but in a lower position. They calibrated the model based on the results of 4,200 previous placements and tested the model on a test data set of 1,200 placements. The performance of the model on the test data set is shown in the confusion matrix below: Actual Predicted Total Success 780 102 882 Failure 74 244 318 854 346 1,200 Success Failure Total Question 1: What is the false positive and true positive rate of the model? Question 2: What is the total error rate of the model? Question 3: A software development company estimates that the cost per hiring failure is $45,000 in lost productivity, severance pay, cost of hiring a replacement, etc. If the company only hired candidates that the model predicted would be successful, what is the estimated average "failure cost" per hire?

Explanation / Answer

QUESTION:

Its a problem on confusion matrix.

False posiitve rate=false positive/true negative+false positive

=74/780+74

=0.08665

0.08665*100=8.67%

Question2:

error rate=1-accuracy

=1-(tp+tn/total)

=1-(780+244/1200)

=1-0.85333

=0.1467

=0.1467*100

=14.67%

Related Questions

Navigate

• Browse (All)
• Browse Q
• Subjects

---

---

Integrity-first tutoring: explanations and feedback only — we do not complete graded work. Learn more.