AIF-C01 - Fundamentals of AI and ML (20% of the exam) - Section 1.3
Describe the machine learning development lifecycle.
Walk the lifecycle from data collection and preparation, through training and evaluation, to deployment and monitoring. Understand why a model degrades over time and why monitoring and retraining are part of the lifecycle, not an afterthought.
ML lifecycleModel monitoringData drift
Practice question for this objective
lock_openFree sampleFundamentals of AI and MLmedium
A retailer deployed a sales-forecasting model a year ago and has not changed its code or weights, yet its accuracy on live traffic has fallen steadily. The team wants to identify genuine reasons a stable, unchanged model degrades in production over time. Which two factors most plausibly explain this kind of gradual decline? Select TWO.
AThe statistical distribution of incoming feature values has drifted away from the distribution the model was trained on.check_circle Correct
BThe relationship between the input features and the target outcome has shifted, so the same inputs now map to different results.check_circle Correct
CThe model's learned parameters slowly decay on their own each month while the endpoint is serving requests.
DThe hardware accelerator running inference rounds floating-point numbers differently as the server ages physically.
EThe training dataset stored in object storage shrinks automatically, leaving the model with less to predict from.
Recognise that data drift and concept drift, not the model artefact itself, are the genuine causes of a deployed model degrading over time. A deployed model's weights are fixed, so degradation comes from the world changing, not the artefact. Data drift shifts the input distribution and concept drift shifts the input-to-output relationship; both leave a frozen model worse aligned with current reality even though its code is unchanged.
Why A is correct: Data drift means live inputs no longer resemble the training data, so a fixed model fits them less well and accuracy slips.
Why B is correct: Concept drift changes the underlying input-to-output mapping the model learned, degrading a model whose weights stay frozen.
Why C is wrong: Tempting because performance falls over time, but stored weights are static files and do not decay or change while serving.
Why D is wrong: Sounds technical, but inference arithmetic does not drift with hardware age, so this cannot explain a steady accuracy loss.
Why E is wrong: Misleading because data matters, yet the training set is not consulted at inference and stored data does not shrink by itself.
Answers that look right on this material and are not. Each one is a distractor from a different question in the AIF-C01 bank for this domain.
Train the model, frame the business problem, deploy the model, collect and prepare data, evaluate the model, then monitor the model in production.
errorWhy it is wrong: This is tempting if a learner fixates on training as the central activity, but you cannot train before defining the problem or preparing data, so beginning with training inverts the dependencies.
Model evaluation, where a held-out test set is used to measure how well the model handles missing inputs at inference time.
errorWhy it is wrong: Tempting because evaluation does expose how a trained model handles real data imperfections, but missing-value imputation and feature scaling must occur before training, not after. Evaluation is a downstream phase that assumes clean, pre-processed inputs.
Choosing the business objective and deciding whether machine learning is even the appropriate approach for the task.
errorWhy it is wrong: Plausible as lifecycle work, but framing the problem happens before any model exists, not after deployment to production.