Ultimate access to all questions.
Your team has developed a custom TensorFlow model designed to identify diagnostic images that require further analysis and medical support. The model demonstrated high accuracy during the testing phase using a comprehensive dataset. However, upon deployment in a real-world hospital setting, the medical staff reports significant dissatisfaction with the model's performance, noting that it frequently misses critical cases that require immediate attention. The hospital's IT infrastructure supports the model's computational requirements, and the data preprocessing pipeline is consistent with the testing phase. Given these constraints, what is the most likely reason for the discrepancy between the model's testing performance and its real-world application? Choose the best option.
A
The model's classification threshold is set too high, causing it to overlook cases that are borderline but still require attention.
B
The deep neural network (DNN) model is suffering from overfitting, having learned the training data too well, including its noise and outliers, which compromises its generalization to unseen data.
C
The deep neural network (DNN) model is suffering from underfitting, failing to capture the underlying patterns in the training data, which affects its performance on both seen and unseen data.
D
Feature crosses need to be performed to improve the model's ability to capture complex relationships in the data, which were not adequately represented during the training phase.