AIF-C01 Question 178
Single answerA healthcare company is using a machine learning model to predict the likelihood of patients developing a chronic disease. During testing, it is observed that the model's predictions are highly accurate for younger patients but significantly less accurate for older patients. What is the most likely reason for this issue?
- A
The model is underfitting due to high bias, resulting in poor performance across all groups.
- B
The training dataset has an imbalance, with fewer samples representing older patients.
- C
The model is overfitting due to high variance, causing it to perform poorly on unseen data.
- D
The model is biased against older patients due to an inherent algorithmic limitation.
Show answer and explanation
Correct answer: B
Explanation
The issue arises because the training dataset has fewer samples representing older patients, leading to the model being less effective at learning patterns for that group. This is a common example of dataset bias, where underrepresentation of a specific demographic group leads to reduced model accuracy for that group.
- A. Incorrect.
Underfitting occurs when the model is too simplistic to capture patterns in the data. This would result in poor performance across all groups, not just one specific demographic group.
- B. Correct.
If the training dataset is imbalanced and contains fewer samples from certain groups (in this case, older patients), the model will likely not perform well for those underrepresented groups.
- C. Incorrect.
Overfitting occurs when a model learns the training data too well, including noise, leading to poor generalization. However, this does not explain why performance is specifically worse for one demographic group.
- D. Incorrect.
Algorithmic bias can exist in models, but it is typically introduced due to imbalanced or biased training data rather than the algorithm itself. The issue here is more likely related to the dataset used.