AIF-C01 Question 177
Select 2A company is deploying a machine learning model to predict loan approvals. During testing, they notice that the model performs exceptionally well on the training data but poorly on the validation data. Additionally, the model tends to deny loans disproportionately to a specific demographic group. What is most likely causing these issues?
- A
The model is overfitting the training data.
- B
The dataset used for training has a bias against the specific demographic group.
- C
The model has high variance but low bias.
- D
The model is underfitting the training data.
- E
The model has high bias but low variance.
Show answer and explanation
Correct answers: A, B
Explanation
The model's poor performance on validation data indicates overfitting, which occurs when the model is too complex and captures noise in the training data. Furthermore, the disproportionate denial of loans to a specific demographic group suggests the presence of bias in the training dataset. These two issues are separate but can coexist, negatively impacting the fairness and accuracy of the machine learning model.
- A. Correct.
Overfitting occurs when a model learns the training data too well, including noise and irrelevant patterns, which leads to poor generalization on unseen data. This matches the scenario where the model performs well on training data but poorly on validation data.
- B. Correct.
Bias in the dataset can lead to unfair treatment of specific demographic groups, which explains why the model denies loans disproportionately to a certain group.
- C. Incorrect.
While high variance indicates overfitting, this option doesn't directly explain the issue of demographic bias in the predictions, so it is not entirely correct in this context.
- D. Incorrect.
Underfitting occurs when the model fails to capture patterns in the training data, leading to poor performance overall. This scenario describes poor performance on validation data but good performance on training data, indicating overfitting rather than underfitting.
- E. Incorrect.
High bias usually leads to underfitting, which is not the case in this scenario. Additionally, it does not explain the demographic bias issue.