Ultimate access to all questions.
You are a data scientist at a leading magazine publisher, tasked with developing a predictive model to identify customers at risk of canceling their annual subscriptions. Your initial exploratory data analysis reveals a significant class imbalance: 90% of customers renew their subscriptions annually, while only 10% cancel. After training a Neural Network Classifier on this dataset, the model achieves 99% accuracy in predicting cancellations and 82% accuracy in predicting renewals. Considering the publisher's goal to minimize customer churn and the challenges posed by imbalanced datasets, which of the following statements best evaluates the model's performance? Choose the two most accurate options.
A
The model's performance is optimal, as it achieves high accuracy in predicting both renewals and cancellations, exceeding the publisher's expectations.
B
The model's performance is suboptimal because it performs worse than a naive strategy that predicts all customers will renew, given the high renewal rate.
C
The model demonstrates exceptional ability in identifying the minority class (cancellations), a challenging task due to the dataset's imbalance, making it valuable for the publisher's churn reduction goals.
D
The model should prioritize achieving higher accuracy for renewals over cancellations, as renewals represent the majority of the dataset and are more critical for revenue.
E
The model's high accuracy in predicting cancellations is misleading, as it may not generalize well to unseen data due to the small size of the cancellation class.