AMAADOR ACADEMYCertificatesVerify a certificate

MLF

Final check — Machine Learning Foundations: Core Concepts Without the Code

Answer the 24 questions. You need 80% or more to pass. You can retry as many times as you like.

Study the course →

Question 1What best describes the difference between machine learning and classic rule-based software?
Question 2A spam filter was trained last year on labelled emails. Today it checks each incoming message and gives a spam score. What is it doing today?
Question 3A manager asks for an ML project ‘to use AI on our customer data’. What should you ask first?
Question 4Why should an ML project define a simple non-ML baseline?
Question 5An insurer has 50,000 past claims, each marked ‘fraud’ or ‘genuine’ after investigation, and wants to score new claims. Which approach fits?
Question 6A logistics firm wants to predict the delivery time in hours for each parcel. What type of task is this?
Question 7Which statement about logistic regression is correct?
Question 8How are large language models mainly pre-trained?
Question 9In a sensor dataset, missing temperatures were stored as 999. The team plans to train directly. What is the risk?
Question 10A team predicts next month’s sales. They shuffle five years of weekly data randomly into training and test sets. What is the problem?
Question 11What is the test set for?
Question 12A hospital model uses ‘discharge summary code’ to predict whether a patient will be readmitted. The code is written when the patient leaves. Why is this a concern?
Question 13A model reaches 99% accuracy on training data but 70% on validation data. What is the most likely issue?
Question 14Which of these is a hyperparameter rather than a learned parameter?
Question 15During training, the loss jumps up and down and never settles. Which adjustment is most likely to help?
Question 16What does k-fold cross-validation do?
Question 17Out of 1,000 transactions, 10 are fraud. A model labels every transaction ‘not fraud’. What is its recall for fraud?
Question 18Precision answers which question?
Question 19A payment provider’s fraud model blocks payments it flags. Customers complain that many genuine payments are blocked. Which metric needs to improve?
Question 20A new demand-forecasting model has an MAE of 120 units per week. ‘Same as last week’ gives an MAE of 115. What should you conclude?
Question 21A team removes ‘gender’ from a hiring model but keeps first name and the applicant’s sports clubs. What should they do before use?
Question 22After a new law changes how customers can cancel contracts, a churn model’s predictions become much less reliable, although input data looks the same. What is this?
Question 23What are the four functions of the NIST AI Risk Management Framework 1.0?
Question 24A demand model has run in production for a year. Nobody owns it, and nobody compares its forecasts with actual sales. What is the first fix?