Anomaly Detection and Optimization GMLE Practice Questions (Page 2)

Questions 6–10

1. 6foundation · easy

   Which statistical test is specifically designed to detect a single outlier in a univariate dataset that is assumed to be normally distributed?

   • AGrubbs' test
   • BChi-squared test
   • CF-test
   • DStudent's t-test
   Reveal Solution Discussion Select an answer first
2. 7application · medium

   A team is tuning a one-class SVM for anomaly detection. They have a small validation set with known anomalies. The team wants to minimize the number of false positives while still catching most anomalies. Which hyperparameter tuning approach is most appropriate?

   • AGrid search over gamma and nu, selecting the combination with the highest F1-score on the validation set.
   • BBayesian optimization with a focus on maximizing recall.
   • CManual tuning of nu based on the expected anomaly rate.
   • DRandom search over gamma only, using the default nu.
   Reveal Solution Discussion Select an answer first
4. 8application · medium

   A monitoring system tracks the number of failed login attempts per minute for a web application. On a typical day, the rate is stable, but during a distributed denial-of-service attack, the rate spikes for several minutes. The security team wants to detect the attack as early as possible. Which type of anomaly best describes this scenario, and which detection approach is most suitable?

   • APoint anomaly; use a z-score on each minute's count.
   • BCollective anomaly; use a sliding window and compare the distribution of counts in the window to the baseline.
   • CContextual anomaly; use a seasonal decomposition to model the expected pattern.
   • DPoint anomaly; use an isolation forest on individual minute counts.
   Reveal Solution Discussion Select an answer first
5. 9application · medium

   A manufacturing plant monitors the temperature of a furnace during a 12-hour curing cycle. The temperature follows a known daily pattern, but occasional sensor glitches produce single readings that are far outside the expected range. The process engineer wants to flag these glitches without raising alerts for normal startup and cooldown phases. Which statistical approach is most appropriate?

   • AApply Grubbs' test iteratively to remove outliers until no further outliers are detected.
   • BCompute a z-score for each reading against the mean and standard deviation of the entire day's data.
   • CModel the expected temperature as a function of time and flag readings that deviate by more than three standard deviations from the expected value at that time.
   • DFit a Gaussian mixture model to the temperature distribution and flag readings in the smallest component.
   Reveal Solution Discussion Select an answer first
6. 10foundation · easy

   When selecting features for an anomaly detection model, which type of feature is generally most useful?

   • AFeatures that have a large number of missing values
   • BFeatures that are highly correlated with each other
   • CFeatures that are constant across all samples
   • DFeatures that capture meaningful deviations from normal behavior
   Reveal Solution Discussion Select an answer first