Classification: Test Your Knowledge

  1. Increasing a binary classifier’s threshold value is likely to produce which of the following effects?

  2. The dataset that you split into train, test and evaluate sets has 9,998 negative examples and 2 positive examples. The resulting model has an accuracy rate of 99.9%. Can you trust this model based on that accuracy metric?

  3. In general, when precision increases, what happens to recall?

  4. True or False: The points on a binary classification model’s ROC (receiver-operating characteristic) curve closest to (1,1) (the upper-right corner) generally represent the best-performing thresholds for the model

  5. You are evaluating the performance of two binary classification models: Model A and Model B. Model A has an AUC of 0.5. Model B's predictions are made completely randomly. Which of the following statements is true?