Topic 1 Question #1
A large mobile network operating company is building a machine learning model to predict customers who are likely to unsubscribe from the service. The company plans to offer an incentive for these customers as the cost of churn is far greater than the cost of the incentive.
The model produces the following confusion matrix after evaluating on a test dataset of 100 customers:
" target="_blank" rel="nofollow noopener">https://www.examtopics.com/assets/media/exam-media/04145/0000200001.jpg">
Based on the model evaluation results, why is this a viable model for production?
Answer: A
Confusion Matrix Analysis:
True Positives (TP): 10 (Actual churn = Yes, Predicted churn = Yes)
False Positives (FP): 10 (Actual churn = No, Predicted churn = Yes)
True Negatives (TN): 76 (Actual churn = No, Predicted churn = No)
False Negatives (FN): 4 (Actual churn = Yes, Predicted churn = No)
Accuracy Calculation:
Accuracy = TP+TNTotal=10+76100=0.86\frac{TP + TN}{Total} = \frac{10 + 76}{100} = 0.86 or 86%.
Cost Justification:
The problem states that the cost of churn (false negatives) is far greater than the cost of providing incentives (false positives).
Since the false negatives (4) are low compared to true positives, the model effectively captures most churners, minimizing potential losses. The model's 86% accuracy and manageable false positives make it viable.
Why Option A?:
The key consideration is the cost impact. The model balances high accuracy with lower false negatives, which aligns with the company’s goal to prevent churn cost-effectively.
Other Options:
B: Precision is not mentioned as 86%; precision = TPTP+FP=1010+10=0.5\frac{TP}{TP + FP} = \frac{10}{10 + 10} = 0.5, or 50%.
C: The focus is on false negatives, not false positives, because false negatives represent churners missed by the model.
D: Precision is less than accuracy, so this is incorrect.<