Microsoft Azure AI Fundamentals (AI-900)Describe fundamental principles of machine learning on AzureMedium
A healthcare provider is using a machine learning model to predict the risk of a patient developing a certain disease. The model was trained on a dataset where 95% of patients are healthy and only 5% develop the disease. If the model achieves 95% accuracy by simply predicting 'healthy' for every patient, what concept does this scenario highlight as a potential issue?
- AFeature scaling necessity
- BImbalanced dataset
- CData leakage
- DOverfitting
Show answer & explanationAnswer & explanation
Correct answer: B. Imbalanced dataset
This scenario highlights an imbalanced dataset, where one class (healthy) significantly outnumbers the other (diseased). A model can achieve high accuracy by predicting the majority class, but it performs poorly on the minority class, rendering the accuracy metric misleading.
Why the other options are wrong
- A. Feature scaling is a preprocessing step for numerical features, not a problem highlighted by skewed accuracy in imbalanced classification.
- C. Data leakage happens when information from outside the training data is used to create the model, leading to overly optimistic performance.
- D. Overfitting occurs when a model learns the training data too well, including noise, and performs poorly on unseen data.
Imbalanced Dataset
A dataset where the number of observations for one class (or category) is significantly lower than for other classes.
- Can lead to misleading accuracy metrics.
- Models may prioritize the majority class.
- Requires specialized techniques (e.g., oversampling, undersampling, using different metrics like F1-score, precision, recall) to address.
Memory trick: Imbalanced data 'skews' the 'score' and makes it 'unbalanced'.